Merge branch 'main' of ssh://github.com/hashicorp/consul

.changelog/11742.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+api: Add filtering support to Catalog's List Services (v1/catalog/services)
+```

.changelog/13493.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+cli: Fix Consul kv CLI 'GET' flags 'keys' and 'recurse' to be set together
+```

.changelog/13998.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+connect: expose new tracing configuration on envoy
+```

.changelog/14034.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+cli: When launching a sidecar proxy with `consul connect envoy` or `consul connect proxy`, the `-sidecar-for` service ID argument is now treated as case-insensitive.
+```

.changelog/14238.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+envoy: adds additional Envoy outlier ejection parameters to passive health check configurations.
+```

.changelog/14269.txt

@@ -0,0 +1,3 @@
+```release-note:bugfix
+connect: Fix issue where `auto_config` and `auto_encrypt` could unintentionally enable TLS for gRPC xDS connections.
+```

.changelog/14290.txt

@@ -0,0 +1,3 @@
+```release-note:bugfix
+envoy: validate name before deleting proxy default configurations. 
+```

.changelog/14343.txt

@@ -0,0 +1,4 @@
+```release-note:feature
+ui: Use withCredentials for all HTTP API requests
+```
+

.changelog/14364.txt

@@ -0,0 +1,3 @@
+```release-note:bugfix
+peering: Fix issue preventing deletion and recreation of peerings in TERMINATED state.
+```

.changelog/14373.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+xds: Set `max_ejection_percent` on Envoy's outlier detection to 100% for peered services.
+```

.changelog/14378.txt

@@ -0,0 +1,5 @@
+```release-note:bug
+api: Fix a breaking change caused by renaming `QueryDatacenterOptions` to
+`QueryFailoverOptions`. This adds `QueryDatacenterOptions` back as an alias to
+`QueryFailoverOptions` and marks it as deprecated.
+```

.changelog/14396.txt

@@ -0,0 +1,3 @@
+```release-note:feature
+peering: Add support to failover to services running on cluster peers.
+```

.changelog/14423.txt

@@ -0,0 +1,3 @@
+```release-note:feature
+cli: Adds new subcommands for `peering` workflows. Refer to the [CLI docs](https://www.consul.io/commands/peering) for more information.
+```

.changelog/14429.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+connect: Fixed an issue where intermediate certificates could build up in the root CA because they were never being pruned after expiring.
+``

.changelog/14433.txt

@@ -0,0 +1,3 @@
+```release-note:bug
+checks: If set, use proxy address for automatically added sidecar check instead of service address.
+```

.changelog/_2271.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+snapshot agent: **(Enterprise only)** Add support for path-based addressing when using s3 backend.
+```

.circleci/config.yml

@@ -816,7 +816,7 @@ jobs:
       # Get go binary from workspace
       - attach_workspace:
           at: .
-      # Build the consul-dev image from the already built binary
+      # Build the consul:local image from the already built binary
       - run:
           command: |
             sudo rm -rf /usr/local/go
@@ -887,8 +887,8 @@ jobs:
       - attach_workspace:
           at: .
       - run: *install-gotestsum
-      # Build the consul-dev image from the already built binary
+      # Build the consul:local image from the already built binary
-      - run: docker build -t consul-dev -f ./build-support/docker/Consul-Dev.dockerfile .
+      - run: docker build -t consul:local -f ./build-support/docker/Consul-Dev.dockerfile .
       - run:
           name: Envoy Integration Tests
           command: |
@@ -902,6 +902,7 @@ jobs:
             GOTESTSUM_JUNITFILE: /tmp/test-results/results.xml
             GOTESTSUM_FORMAT: standard-verbose
             COMPOSE_INTERACTIVE_NO_CLI: 1
+            LAMBDA_TESTS_ENABLED: "true"
             # tput complains if this isn't set to something.
             TERM: ansi
       - store_artifacts:

.github/workflows/backport-assistant.yml

@@ -16,7 +16,7 @@ jobs:
   backport:
     if: github.event.pull_request.merged
     runs-on: ubuntu-latest
-    container: hashicorpdev/backport-assistant:0.2.3
+    container: hashicorpdev/backport-assistant:0.2.5
     steps:
       - name: Run Backport Assistant for stable-website
         run: |
@@ -24,6 +24,7 @@ jobs:
         env:
           BACKPORT_LABEL_REGEXP: "type/docs-(?P<target>cherrypick)"
           BACKPORT_TARGET_TEMPLATE: "stable-website"
+          BACKPORT_MERGE_COMMIT: true
           GITHUB_TOKEN: ${{ secrets.ELEVATED_GITHUB_TOKEN }}
       - name: Backport changes to latest release branch
         run: |

.golangci.yml

@@ -8,6 +8,8 @@ linters:
     - ineffassign
     - unparam
     - forbidigo
+    - gomodguard
+    - depguard
 
 issues:
   # Disable the default exclude list so that all excludes are explicitly
@@ -75,6 +77,30 @@ linters-settings:
     # Exclude godoc examples from forbidigo checks.
     # Default: true
     exclude_godoc_examples: false
+  gomodguard:
+    blocked:
+      # List of blocked modules.
+      modules:
+        # Blocked module.
+        - github.com/hashicorp/net-rpc-msgpackrpc:
+            recommendations:
+              - github.com/hashicorp/consul-net-rpc/net-rpc-msgpackrpc
+        - github.com/hashicorp/go-msgpack:
+            recommendations:
+              - github.com/hashicorp/consul-net-rpc/go-msgpack
+
+  depguard:
+    list-type: denylist
+    include-go-root: true
+    # A list of packages for the list type specified.
+    # Default: []
+    packages:
+      - net/rpc
+    # A list of packages for the list type specified.
+    # Specify an error message to output when a denied package is used.
+    # Default: []
+    packages-with-error-message:
+      - net/rpc: 'only use forked copy in github.com/hashicorp/consul-net-rpc/net/rpc'
 
 run:
   timeout: 10m

GNUmakefile

@@ -16,6 +16,7 @@ PROTOC_GO_INJECT_TAG_VERSION='v1.3.0'
 
 GOTAGS ?=
 GOPATH=$(shell go env GOPATH)
+GOARCH?=$(shell go env GOARCH)
 MAIN_GOPATH=$(shell go env GOPATH | cut -d: -f1)
 
 export PATH := $(PWD)/bin:$(GOPATH)/bin:$(PATH)
@@ -129,7 +130,7 @@ export GOLDFLAGS
 
 # Allow skipping docker build during integration tests in CI since we already
 # have a built binary
-ENVOY_INTEG_DEPS?=dev-docker
+ENVOY_INTEG_DEPS?=docker-envoy-integ
 ifdef SKIP_DOCKER_BUILD
 ENVOY_INTEG_DEPS=noop
 endif
@@ -152,7 +153,28 @@ dev-docker: linux
 	@docker pull consul:$(CONSUL_IMAGE_VERSION) >/dev/null
 	@echo "Building Consul Development container - $(CONSUL_DEV_IMAGE)"
 	#  'consul:local' tag is needed to run the integration tests
-	@DOCKER_DEFAULT_PLATFORM=linux/amd64 docker build $(NOCACHE) $(QUIET) -t '$(CONSUL_DEV_IMAGE)' -t 'consul:local' --build-arg CONSUL_IMAGE_VERSION=$(CONSUL_IMAGE_VERSION) $(CURDIR)/pkg/bin/linux_amd64 -f $(CURDIR)/build-support/docker/Consul-Dev.dockerfile
+	@docker buildx use default && docker buildx build -t 'consul:local' \
+       --platform linux/$(GOARCH) \
+	   --build-arg CONSUL_IMAGE_VERSION=$(CONSUL_IMAGE_VERSION) \
+       --load \
+       -f $(CURDIR)/build-support/docker/Consul-Dev-Multiarch.dockerfile $(CURDIR)/pkg/bin/
+
+check-remote-dev-image-env:
+ifndef REMOTE_DEV_IMAGE
+	$(error REMOTE_DEV_IMAGE is undefined: set this image to <your_docker_repo>/<your_docker_image>:<image_tag>, e.g. hashicorp/consul-k8s-dev:latest)
+endif
+
+remote-docker: check-remote-dev-image-env
+	$(MAKE) GOARCH=amd64 linux
+	$(MAKE) GOARCH=arm64 linux
+	@echo "Pulling consul container image - $(CONSUL_IMAGE_VERSION)"
+	@docker pull consul:$(CONSUL_IMAGE_VERSION) >/dev/null
+	@echo "Building and Pushing Consul Development container - $(REMOTE_DEV_IMAGE)"
+	@docker buildx use default && docker buildx build -t '$(REMOTE_DEV_IMAGE)' \
+       --platform linux/amd64,linux/arm64 \
+	   --build-arg CONSUL_IMAGE_VERSION=$(CONSUL_IMAGE_VERSION) \
+       --push \
+       -f $(CURDIR)/build-support/docker/Consul-Dev-Multiarch.dockerfile $(CURDIR)/pkg/bin/
 
 # In CircleCI, the linux binary will be attached from a previous step at bin/. This make target
 # should only run in CI and not locally.
@@ -174,10 +196,10 @@ ifeq ($(CIRCLE_BRANCH), main)
 	@docker push $(CI_DEV_DOCKER_NAMESPACE)/$(CI_DEV_DOCKER_IMAGE_NAME):latest
 endif
 
-# linux builds a linux binary independent of the source platform
+# linux builds a linux binary compatible with the source platform
 linux:
-	@mkdir -p ./pkg/bin/linux_amd64
+	@mkdir -p ./pkg/bin/linux_$(GOARCH)
-	CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -o ./pkg/bin/linux_amd64 -ldflags "$(GOLDFLAGS)" -tags "$(GOTAGS)"
+	CGO_ENABLED=0 GOOS=linux GOARCH=$(GOARCH) go build -o ./pkg/bin/linux_$(GOARCH) -ldflags "$(GOLDFLAGS)" -tags "$(GOTAGS)"
 
 # dist builds binaries for all platforms and packages them for distribution
 dist:
@@ -324,8 +346,22 @@ consul-docker: go-build-image
 ui-docker: ui-build-image
 	@$(SHELL) $(CURDIR)/build-support/scripts/build-docker.sh ui
 
+# Build image used to run integration tests locally.
+docker-envoy-integ:
+	$(MAKE) GOARCH=amd64 linux
+	docker build \
+      --platform linux/amd64 $(NOCACHE) $(QUIET) \
+      -t 'consul:local' \
+      --build-arg CONSUL_IMAGE_VERSION=$(CONSUL_IMAGE_VERSION) \
+      $(CURDIR)/pkg/bin/linux_amd64 \
+      -f $(CURDIR)/build-support/docker/Consul-Dev.dockerfile
+
+# Run integration tests.
+# Use GO_TEST_FLAGS to run specific tests:
+#      make test-envoy-integ GO_TEST_FLAGS="-run TestEnvoy/case-basic"
+# NOTE: Always uses amd64 images, even when running on M1 macs, to match CI/CD environment.
 test-envoy-integ: $(ENVOY_INTEG_DEPS)
-	@go test -v -timeout=30m -tags integration ./test/integration/connect/envoy
+	@go test -v -timeout=30m -tags integration $(GO_TEST_FLAGS) ./test/integration/connect/envoy
 
 .PHONY: test-compat-integ
 test-compat-integ: dev-docker

agent/agent_endpoint_test.go

@@ -3764,7 +3764,7 @@ func testAgent_RegisterService_TranslateKeys(t *testing.T, extraHCL string) {
 				fmt.Println("TCP Check:= ", v)
 			}
 			if hasNoCorrectTCPCheck {
-				t.Fatalf("Did not find the expected TCP Healtcheck '%s' in %#v ", tt.expectedTCPCheckStart, a.checkTCPs)
+				t.Fatalf("Did not find the expected TCP Healthcheck '%s' in %#v ", tt.expectedTCPCheckStart, a.checkTCPs)
 			}
 			require.Equal(t, sidecarSvc, gotSidecar)
 		})

agent/config/builder.go

@@ -2531,10 +2531,9 @@ func (b *builder) buildTLSConfig(rt RuntimeConfig, t TLS) (tlsutil.Config, error
 		return c, errors.New("verify_server_hostname is only valid in the tls.internal_rpc stanza")
 	}
 
-	// TLS is only enabled on the gRPC listener if there's an HTTPS port configured
+	// And UseAutoCert right now only applies to external gRPC interface.
-	// for historic and backwards-compatibility reasons.
+	if t.Defaults.UseAutoCert != nil || t.HTTPS.UseAutoCert != nil || t.InternalRPC.UseAutoCert != nil {
-	if rt.HTTPSPort <= 0 && (t.GRPC != TLSProtocolConfig{} && t.GRPCModifiedByDeprecatedConfig == nil) {
+		return c, errors.New("use_auto_cert is only valid in the tls.grpc stanza")
-		b.warn("tls.grpc was provided but TLS will NOT be enabled on the gRPC listener without an HTTPS listener configured (e.g. via ports.https)")
 	}
 
 	defaultTLSMinVersion := b.tlsVersion("tls.defaults.tls_min_version", t.Defaults.TLSMinVersion)
@@ -2591,6 +2590,7 @@ func (b *builder) buildTLSConfig(rt RuntimeConfig, t TLS) (tlsutil.Config, error
 
 	mapCommon("https", t.HTTPS, &c.HTTPS)
 	mapCommon("grpc", t.GRPC, &c.GRPC)
+	c.GRPC.UseAutoCert = boolValWithDefault(t.GRPC.UseAutoCert, false)
 
 	c.ServerName = rt.ServerName
 	c.NodeName = rt.NodeName

agent/config/config.go

@@ -867,6 +867,7 @@ type TLSProtocolConfig struct {
 	VerifyIncoming       *bool   `mapstructure:"verify_incoming"`
 	VerifyOutgoing       *bool   `mapstructure:"verify_outgoing"`
 	VerifyServerHostname *bool   `mapstructure:"verify_server_hostname"`
+	UseAutoCert          *bool   `mapstructure:"use_auto_cert"`
 }
 
 type TLS struct {

agent/config/runtime_test.go

@@ -5516,7 +5516,70 @@ func TestLoad_IntegrationWithFlags(t *testing.T) {
 		},
 	})
 	run(t, testCase{
-		desc: "tls.grpc without ports.https",
+		desc: "tls.grpc.use_auto_cert defaults to false",
+		args: []string{
+			`-data-dir=` + dataDir,
+		},
+		json: []string{`
+			{
+				"tls": {
+					"grpc": {}
+				}
+			}
+		`},
+		hcl: []string{`
+			tls {
+				grpc {}
+			}
+		`},
+		expected: func(rt *RuntimeConfig) {
+			rt.DataDir = dataDir
+			rt.TLS.Domain = "consul."
+			rt.TLS.NodeName = "thehostname"
+			rt.TLS.GRPC.UseAutoCert = false
+		},
+	})
+	run(t, testCase{
+		desc: "tls.grpc.use_auto_cert defaults to false (II)",
+		args: []string{
+			`-data-dir=` + dataDir,
+		},
+		json: []string{`
+			{
+				"tls": {}
+			}
+		`},
+		hcl: []string{`
+			tls {
+			}
+		`},
+		expected: func(rt *RuntimeConfig) {
+			rt.DataDir = dataDir
+			rt.TLS.Domain = "consul."
+			rt.TLS.NodeName = "thehostname"
+			rt.TLS.GRPC.UseAutoCert = false
+		},
+	})
+	run(t, testCase{
+		desc: "tls.grpc.use_auto_cert defaults to false (III)",
+		args: []string{
+			`-data-dir=` + dataDir,
+		},
+		json: []string{`
+			{
+			}
+		`},
+		hcl: []string{`
+		`},
+		expected: func(rt *RuntimeConfig) {
+			rt.DataDir = dataDir
+			rt.TLS.Domain = "consul."
+			rt.TLS.NodeName = "thehostname"
+			rt.TLS.GRPC.UseAutoCert = false
+		},
+	})
+	run(t, testCase{
+		desc: "tls.grpc.use_auto_cert enabled when true",
 		args: []string{
 			`-data-dir=` + dataDir,
 		},
@@ -5524,7 +5587,7 @@ func TestLoad_IntegrationWithFlags(t *testing.T) {
 			{
 				"tls": {
 					"grpc": {
-						"cert_file": "cert-1234"
+						"use_auto_cert": true
 					}
 				}
 			}
@@ -5532,20 +5595,43 @@ func TestLoad_IntegrationWithFlags(t *testing.T) {
 		hcl: []string{`
 			tls {
 				grpc {
-					cert_file = "cert-1234"
+					use_auto_cert = true
 				}
 			}
 		`},
 		expected: func(rt *RuntimeConfig) {
 			rt.DataDir = dataDir
-
 			rt.TLS.Domain = "consul."
 			rt.TLS.NodeName = "thehostname"
-
+			rt.TLS.GRPC.UseAutoCert = true
-			rt.TLS.GRPC.CertFile = "cert-1234"
 		},
-		expectedWarnings: []string{
+	})
-			"tls.grpc was provided but TLS will NOT be enabled on the gRPC listener without an HTTPS listener configured (e.g. via ports.https)",
+	run(t, testCase{
+		desc: "tls.grpc.use_auto_cert disabled when false",
+		args: []string{
+			`-data-dir=` + dataDir,
+		},
+		json: []string{`
+			{
+				"tls": {
+					"grpc": {
+						"use_auto_cert": false
+					}
+				}
+			}
+		`},
+		hcl: []string{`
+			tls {
+				grpc {
+					use_auto_cert = false
+				}
+			}
+		`},
+		expected: func(rt *RuntimeConfig) {
+			rt.DataDir = dataDir
+			rt.TLS.Domain = "consul."
+			rt.TLS.NodeName = "thehostname"
+			rt.TLS.GRPC.UseAutoCert = false
 		},
 	})
 }
@@ -6340,6 +6426,7 @@ func TestLoad_FullConfig(t *testing.T) {
 				TLSMinVersion:  types.TLSv1_0,
 				CipherSuites:   []types.TLSCipherSuite{types.TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256, types.TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA},
 				VerifyOutgoing: false,
+				UseAutoCert:    true,
 			},
 			HTTPS: tlsutil.ProtocolConfig{
 				VerifyIncoming: true,

agent/config/testdata/TestRuntimeConfig_Sanitize.golden

@@ -374,7 +374,8 @@
             "TLSMinVersion": "",
             "VerifyIncoming": false,
             "VerifyOutgoing": false,
-            "VerifyServerHostname": false
+            "VerifyServerHostname": false,
+            "UseAutoCert": false
         },
         "HTTPS": {
             "CAFile": "",
@@ -385,7 +386,8 @@
             "TLSMinVersion": "",
             "VerifyIncoming": false,
             "VerifyOutgoing": false,
-            "VerifyServerHostname": false
+            "VerifyServerHostname": false,
+            "UseAutoCert": false
         },
         "InternalRPC": {
             "CAFile": "",
@@ -396,7 +398,8 @@
             "TLSMinVersion": "",
             "VerifyIncoming": false,
             "VerifyOutgoing": false,
-            "VerifyServerHostname": false
+            "VerifyServerHostname": false,
+            "UseAutoCert": false
         },
         "NodeName": "",
         "ServerName": ""
@@ -466,4 +469,4 @@
     "VersionMetadata": "",
     "VersionPrerelease": "",
     "Watches": []
-}
+}

agent/config/testdata/full-config.hcl

@@ -697,6 +697,7 @@ tls {
         tls_cipher_suites = "TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256,TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA"
         tls_min_version = "TLSv1_0"
         verify_incoming = true
+        use_auto_cert   = true
     }
 }
 tls_cipher_suites = "TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA,TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256"

agent/config/testdata/full-config.json

@@ -692,7 +692,8 @@
       "key_file": "1y4prKjl",
       "tls_cipher_suites": "TLS_ECDHE_ECDSA_WITH_CHACHA20_POLY1305_SHA256,TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA",
       "tls_min_version": "TLSv1_0",
-      "verify_incoming": true
+      "verify_incoming": true,
+      "use_auto_cert": true
     }
   },
   "tls_cipher_suites": "TLS_ECDHE_ECDSA_WITH_AES_128_CBC_SHA,TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256",

agent/connect/sni_test.go

@@ -178,20 +178,43 @@ func TestQuerySNI(t *testing.T) {
 func TestTargetSNI(t *testing.T) {
 	// empty namespace, empty subset
 	require.Equal(t, "api.default.foo."+testTrustDomainSuffix1,
-		TargetSNI(structs.NewDiscoveryTarget("api", "", "", "default", "foo"), testTrustDomain1))
+		TargetSNI(structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:    "api",
+			Partition:  "default",
+			Datacenter: "foo",
+		}), testTrustDomain1))
 
 	require.Equal(t, "api.default.foo."+testTrustDomainSuffix1,
-		TargetSNI(structs.NewDiscoveryTarget("api", "", "", "", "foo"), testTrustDomain1))
+		TargetSNI(structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:    "api",
+			Datacenter: "foo",
+		}), testTrustDomain1))
 
 	// set namespace, empty subset
 	require.Equal(t, "api.neighbor.foo."+testTrustDomainSuffix2,
-		TargetSNI(structs.NewDiscoveryTarget("api", "", "neighbor", "default", "foo"), testTrustDomain2))
+		TargetSNI(structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:    "api",
+			Namespace:  "neighbor",
+			Partition:  "default",
+			Datacenter: "foo",
+		}), testTrustDomain2))
 
 	// empty namespace, set subset
 	require.Equal(t, "v2.api.default.foo."+testTrustDomainSuffix1,
-		TargetSNI(structs.NewDiscoveryTarget("api", "v2", "", "default", "foo"), testTrustDomain1))
+		TargetSNI(structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:       "api",
+			ServiceSubset: "v2",
+			Partition:     "default",
+			Datacenter:    "foo",
+		}), testTrustDomain1))
 
 	// set namespace, set subset
 	require.Equal(t, "canary.api.neighbor.foo."+testTrustDomainSuffix2,
-		TargetSNI(structs.NewDiscoveryTarget("api", "canary", "neighbor", "default", "foo"), testTrustDomain2))
+		TargetSNI(structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:       "api",
+			ServiceSubset: "canary",
+			Namespace:     "neighbor",
+			Partition:     "default",
+			Datacenter:    "foo",
+		}), testTrustDomain2))
 }

agent/consul/catalog_endpoint.go

@@ -565,6 +565,11 @@ func (c *Catalog) ListServices(args *structs.DCSpecificRequest, reply *structs.I
 		return err
 	}
 
+	filter, err := bexpr.CreateFilter(args.Filter, nil, []*structs.ServiceNode{})
+	if err != nil {
+		return err
+	}
+
 	// Set reply enterprise metadata after resolving and validating the token so
 	// that we can properly infer metadata from the token.
 	reply.EnterpriseMeta = args.EnterpriseMeta
@@ -574,10 +579,11 @@ func (c *Catalog) ListServices(args *structs.DCSpecificRequest, reply *structs.I
 		&reply.QueryMeta,
 		func(ws memdb.WatchSet, state *state.Store) error {
 			var err error
+			var serviceNodes structs.ServiceNodes
 			if len(args.NodeMetaFilters) > 0 {
-				reply.Index, reply.Services, err = state.ServicesByNodeMeta(ws, args.NodeMetaFilters, &args.EnterpriseMeta, args.PeerName)
+				reply.Index, serviceNodes, err = state.ServicesByNodeMeta(ws, args.NodeMetaFilters, &args.EnterpriseMeta, args.PeerName)
 			} else {
-				reply.Index, reply.Services, err = state.Services(ws, &args.EnterpriseMeta, args.PeerName)
+				reply.Index, serviceNodes, err = state.Services(ws, &args.EnterpriseMeta, args.PeerName)
 			}
 			if err != nil {
 				return err
@@ -588,11 +594,43 @@ func (c *Catalog) ListServices(args *structs.DCSpecificRequest, reply *structs.I
 				return nil
 			}
 
+			raw, err := filter.Execute(serviceNodes)
+			if err != nil {
+				return err
+			}
+
+			reply.Services = servicesTagsByName(raw.(structs.ServiceNodes))
+
 			c.srv.filterACLWithAuthorizer(authz, reply)
+
 			return nil
 		})
 }
 
+func servicesTagsByName(services []*structs.ServiceNode) structs.Services {
+	unique := make(map[string]map[string]struct{})
+	for _, svc := range services {
+		tags, ok := unique[svc.ServiceName]
+		if !ok {
+			unique[svc.ServiceName] = make(map[string]struct{})
+			tags = unique[svc.ServiceName]
+		}
+		for _, tag := range svc.ServiceTags {
+			tags[tag] = struct{}{}
+		}
+	}
+
+	// Generate the output structure.
+	var results = make(structs.Services)
+	for service, tags := range unique {
+		results[service] = make([]string, 0, len(tags))
+		for tag := range tags {
+			results[service] = append(results[service], tag)
+		}
+	}
+	return results
+}
+
 // ServiceList is used to query the services in a DC.
 // Returns services as a list of ServiceNames.
 func (c *Catalog) ServiceList(args *structs.DCSpecificRequest, reply *structs.IndexedServiceList) error {

agent/consul/catalog_endpoint_test.go

@@ -1523,6 +1523,45 @@ func TestCatalog_ListServices_NodeMetaFilter(t *testing.T) {
 	}
 }
 
+func TestCatalog_ListServices_Filter(t *testing.T) {
+	t.Parallel()
+	_, s1 := testServer(t)
+	codec := rpcClient(t, s1)
+
+	testrpc.WaitForTestAgent(t, s1.RPC, "dc1")
+
+	// prep the cluster with some data we can use in our filters
+	registerTestCatalogEntries(t, codec)
+
+	// Run the tests against the test server
+
+	t.Run("ListServices", func(t *testing.T) {
+		args := structs.DCSpecificRequest{
+			Datacenter: "dc1",
+		}
+
+		args.Filter = "ServiceName == redis"
+		out := new(structs.IndexedServices)
+		require.NoError(t, msgpackrpc.CallWithCodec(codec, "Catalog.ListServices", &args, out))
+		require.Contains(t, out.Services, "redis")
+		require.ElementsMatch(t, []string{"v1", "v2"}, out.Services["redis"])
+
+		args.Filter = "NodeMeta.os == NoSuchOS"
+		out = new(structs.IndexedServices)
+		require.NoError(t, msgpackrpc.CallWithCodec(codec, "Catalog.ListServices", &args, out))
+		require.Len(t, out.Services, 0)
+
+		args.Filter = "NodeMeta.NoSuchMetadata == linux"
+		out = new(structs.IndexedServices)
+		require.NoError(t, msgpackrpc.CallWithCodec(codec, "Catalog.ListServices", &args, out))
+		require.Len(t, out.Services, 0)
+
+		args.Filter = "InvalidField == linux"
+		out = new(structs.IndexedServices)
+		require.Error(t, msgpackrpc.CallWithCodec(codec, "Catalog.ListServices", &args, out))
+	})
+}
+
 func TestCatalog_ListServices_Blocking(t *testing.T) {
 	if testing.Short() {
 		t.Skip("too slow for testing.Short")

agent/consul/config_endpoint_test.go

@@ -1399,8 +1399,9 @@ func TestConfigEntry_ResolveServiceConfig_Upstreams(t *testing.T) {
 							Protocol:    "http",
 							MeshGateway: structs.MeshGatewayConfig{Mode: structs.MeshGatewayModeRemote},
 							PassiveHealthCheck: &structs.PassiveHealthCheck{
-								Interval:    10,
+								Interval:                10,
-								MaxFailures: 2,
+								MaxFailures:             2,
+								EnforcingConsecutive5xx: uintPointer(60),
 							},
 						},
 						Overrides: []*structs.UpstreamConfig{
@@ -1432,8 +1433,9 @@ func TestConfigEntry_ResolveServiceConfig_Upstreams(t *testing.T) {
 						Upstream: wildcard,
 						Config: map[string]interface{}{
 							"passive_health_check": map[string]interface{}{
-								"Interval":    int64(10),
+								"Interval":                int64(10),
-								"MaxFailures": int64(2),
+								"MaxFailures":             int64(2),
+								"EnforcingConsecutive5xx": int64(60),
 							},
 							"mesh_gateway": map[string]interface{}{
 								"Mode": "remote",
@@ -1445,8 +1447,9 @@ func TestConfigEntry_ResolveServiceConfig_Upstreams(t *testing.T) {
 						Upstream: mysql,
 						Config: map[string]interface{}{
 							"passive_health_check": map[string]interface{}{
-								"Interval":    int64(10),
+								"Interval":                int64(10),
-								"MaxFailures": int64(2),
+								"MaxFailures":             int64(2),
+								"EnforcingConsecutive5xx": int64(60),
 							},
 							"mesh_gateway": map[string]interface{}{
 								"Mode": "local",
@@ -2507,3 +2510,7 @@ func Test_gateWriteToSecondary_AllowedKinds(t *testing.T) {
 		})
 	}
 }
+
+func uintPointer(v uint32) *uint32 {
+	return &v
+}

agent/consul/discovery_chain_endpoint_test.go

@@ -56,8 +56,17 @@ func TestDiscoveryChainEndpoint_Get(t *testing.T) {
 		return &resp, nil
 	}
 
-	newTarget := func(service, serviceSubset, namespace, partition, datacenter string) *structs.DiscoveryTarget {
+	newTarget := func(opts structs.DiscoveryTargetOpts) *structs.DiscoveryTarget {
-		t := structs.NewDiscoveryTarget(service, serviceSubset, namespace, partition, datacenter)
+		if opts.Namespace == "" {
+			opts.Namespace = "default"
+		}
+		if opts.Partition == "" {
+			opts.Partition = "default"
+		}
+		if opts.Datacenter == "" {
+			opts.Datacenter = "dc1"
+		}
+		t := structs.NewDiscoveryTarget(opts)
 		t.SNI = connect.TargetSNI(t, connect.TestClusterID+".consul")
 		t.Name = t.SNI
 		t.ConnectTimeout = 5 * time.Second // default
@@ -119,7 +128,7 @@ func TestDiscoveryChainEndpoint_Get(t *testing.T) {
 				},
 			},
 			Targets: map[string]*structs.DiscoveryTarget{
-				"web.default.default.dc1": newTarget("web", "", "default", "default", "dc1"),
+				"web.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 			},
 		},
 	}
@@ -245,7 +254,7 @@ func TestDiscoveryChainEndpoint_Get(t *testing.T) {
 				},
 				Targets: map[string]*structs.DiscoveryTarget{
 					"web.default.default.dc1": targetWithConnectTimeout(
-						newTarget("web", "", "default", "default", "dc1"),
+						newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 						33*time.Second,
 					),
 				},

agent/consul/discoverychain/compile.go

@@ -8,6 +8,7 @@ import (
 	"github.com/mitchellh/hashstructure"
 	"github.com/mitchellh/mapstructure"
 
+	"github.com/hashicorp/consul/acl"
 	"github.com/hashicorp/consul/agent/configentry"
 	"github.com/hashicorp/consul/agent/connect"
 	"github.com/hashicorp/consul/agent/structs"
@@ -576,7 +577,10 @@ func (c *compiler) assembleChain() error {
 	if router == nil {
 		// If no router is configured, move on down the line to the next hop of
 		// the chain.
-		node, err := c.getSplitterOrResolverNode(c.newTarget(c.serviceName, "", "", "", ""))
+		node, err := c.getSplitterOrResolverNode(c.newTarget(structs.DiscoveryTargetOpts{
+			Service: c.serviceName,
+		}))
+
 		if err != nil {
 			return err
 		}
@@ -626,11 +630,20 @@ func (c *compiler) assembleChain() error {
 		)
 		if dest.ServiceSubset == "" {
 			node, err = c.getSplitterOrResolverNode(
-				c.newTarget(svc, "", destNamespace, destPartition, ""),
+				c.newTarget(structs.DiscoveryTargetOpts{
-			)
+					Service:   svc,
+					Namespace: destNamespace,
+					Partition: destPartition,
+				},
+				))
 		} else {
 			node, err = c.getResolverNode(
-				c.newTarget(svc, dest.ServiceSubset, destNamespace, destPartition, ""),
+				c.newTarget(structs.DiscoveryTargetOpts{
+					Service:       svc,
+					ServiceSubset: dest.ServiceSubset,
+					Namespace:     destNamespace,
+					Partition:     destPartition,
+				}),
 				false,
 			)
 		}
@@ -642,7 +655,12 @@ func (c *compiler) assembleChain() error {
 
 	// If we have a router, we'll add a catch-all route at the end to send
 	// unmatched traffic to the next hop in the chain.
-	defaultDestinationNode, err := c.getSplitterOrResolverNode(c.newTarget(router.Name, "", router.NamespaceOrDefault(), router.PartitionOrDefault(), ""))
+	opts := structs.DiscoveryTargetOpts{
+		Service:   router.Name,
+		Namespace: router.NamespaceOrDefault(),
+		Partition: router.PartitionOrDefault(),
+	}
+	defaultDestinationNode, err := c.getSplitterOrResolverNode(c.newTarget(opts))
 	if err != nil {
 		return err
 	}
@@ -674,26 +692,36 @@ func newDefaultServiceRoute(serviceName, namespace, partition string) *structs.S
 	}
 }
 
-func (c *compiler) newTarget(service, serviceSubset, namespace, partition, datacenter string) *structs.DiscoveryTarget {
+func (c *compiler) newTarget(opts structs.DiscoveryTargetOpts) *structs.DiscoveryTarget {
-	if service == "" {
+	if opts.Service == "" {
 		panic("newTarget called with empty service which makes no sense")
 	}
 
-	t := structs.NewDiscoveryTarget(
+	if opts.Peer == "" {
-		service,
+		opts.Datacenter = defaultIfEmpty(opts.Datacenter, c.evaluateInDatacenter)
-		serviceSubset,
+		opts.Namespace = defaultIfEmpty(opts.Namespace, c.evaluateInNamespace)
-		defaultIfEmpty(namespace, c.evaluateInNamespace),
+		opts.Partition = defaultIfEmpty(opts.Partition, c.evaluateInPartition)
-		defaultIfEmpty(partition, c.evaluateInPartition),
+	} else {
-		defaultIfEmpty(datacenter, c.evaluateInDatacenter),
+		// Don't allow Peer and Datacenter.
-	)
+		opts.Datacenter = ""
+		// Peer and Partition cannot both be set.
+		opts.Partition = acl.PartitionOrDefault("")
+		// Default to "default" rather than c.evaluateInNamespace.
+		opts.Namespace = acl.PartitionOrDefault(opts.Namespace)
+	}
 
-	// Set default connect SNI. This will be overridden later if the service
+	t := structs.NewDiscoveryTarget(opts)
-	// has an explicit SNI value configured in service-defaults.
-	t.SNI = connect.TargetSNI(t, c.evaluateInTrustDomain)
 
-	// Use the same representation for the name. This will NOT be overridden
+	// We don't have the peer's trust domain yet so we can't construct the SNI.
-	// later.
+	if opts.Peer == "" {
-	t.Name = t.SNI
+		// Set default connect SNI. This will be overridden later if the service
+		// has an explicit SNI value configured in service-defaults.
+		t.SNI = connect.TargetSNI(t, c.evaluateInTrustDomain)
+
+		// Use the same representation for the name. This will NOT be overridden
+		// later.
+		t.Name = t.SNI
+	}
 
 	prev, ok := c.loadedTargets[t.ID]
 	if ok {
@@ -703,34 +731,30 @@ func (c *compiler) newTarget(service, serviceSubset, namespace, partition, datac
 	return t
 }
 
-func (c *compiler) rewriteTarget(t *structs.DiscoveryTarget, service, serviceSubset, partition, namespace, datacenter string) *structs.DiscoveryTarget {
+func (c *compiler) rewriteTarget(t *structs.DiscoveryTarget, opts structs.DiscoveryTargetOpts) *structs.DiscoveryTarget {
-	var (
+	mergedOpts := t.ToDiscoveryTargetOpts()
-		service2       = t.Service
-		serviceSubset2 = t.ServiceSubset
-		partition2     = t.Partition
-		namespace2     = t.Namespace
-		datacenter2    = t.Datacenter
-	)
 
-	if service != "" && service != service2 {
+	if opts.Service != "" && opts.Service != mergedOpts.Service {
-		service2 = service
+		mergedOpts.Service = opts.Service
 		// Reset the chosen subset if we reference a service other than our own.
-		serviceSubset2 = ""
+		mergedOpts.ServiceSubset = ""
 	}
-	if serviceSubset != "" {
+	if opts.ServiceSubset != "" {
-		serviceSubset2 = serviceSubset
+		mergedOpts.ServiceSubset = opts.ServiceSubset
 	}
-	if partition != "" {
+	if opts.Partition != "" {
-		partition2 = partition
+		mergedOpts.Partition = opts.Partition
 	}
-	if namespace != "" {
+	// Only use explicit Namespace with Peer
-		namespace2 = namespace
+	if opts.Namespace != "" || opts.Peer != "" {
+		mergedOpts.Namespace = opts.Namespace
 	}
-	if datacenter != "" {
+	if opts.Datacenter != "" {
-		datacenter2 = datacenter
+		mergedOpts.Datacenter = opts.Datacenter
 	}
+	mergedOpts.Peer = opts.Peer
 
-	return c.newTarget(service2, serviceSubset2, namespace2, partition2, datacenter2)
+	return c.newTarget(mergedOpts)
 }
 
 func (c *compiler) getSplitterOrResolverNode(target *structs.DiscoveryTarget) (*structs.DiscoveryGraphNode, error) {
@@ -803,10 +827,13 @@ func (c *compiler) getSplitterNode(sid structs.ServiceID) (*structs.DiscoveryGra
 			// fall through to group-resolver
 		}
 
-		node, err := c.getResolverNode(
+		opts := structs.DiscoveryTargetOpts{
-			c.newTarget(splitID.ID, split.ServiceSubset, splitID.NamespaceOrDefault(), splitID.PartitionOrDefault(), ""),
+			Service:       splitID.ID,
-			false,
+			ServiceSubset: split.ServiceSubset,
-		)
+			Namespace:     splitID.NamespaceOrDefault(),
+			Partition:     splitID.PartitionOrDefault(),
+		}
+		node, err := c.getResolverNode(c.newTarget(opts), false)
 		if err != nil {
 			return nil, err
 		}
@@ -881,11 +908,7 @@ RESOLVE_AGAIN:
 
 		redirectedTarget := c.rewriteTarget(
 			target,
-			redirect.Service,
+			redirect.ToDiscoveryTargetOpts(),
-			redirect.ServiceSubset,
-			redirect.Partition,
-			redirect.Namespace,
-			redirect.Datacenter,
 		)
 		if redirectedTarget.ID != target.ID {
 			target = redirectedTarget
@@ -895,14 +918,9 @@ RESOLVE_AGAIN:
 
 	// Handle default subset.
 	if target.ServiceSubset == "" && resolver.DefaultSubset != "" {
-		target = c.rewriteTarget(
+		target = c.rewriteTarget(target, structs.DiscoveryTargetOpts{
-			target,
+			ServiceSubset: resolver.DefaultSubset,
-			"",
+		})
-			resolver.DefaultSubset,
-			"",
-			"",
-			"",
-		)
 		goto RESOLVE_AGAIN
 	}
 
@@ -1027,56 +1045,54 @@ RESOLVE_AGAIN:
 			failover, ok = f["*"]
 		}
 
-		if ok {
+		if !ok {
-			// Determine which failover definitions apply.
+			return node, nil
-			var failoverTargets []*structs.DiscoveryTarget
+		}
-			if len(failover.Datacenters) > 0 {
+
-				for _, dc := range failover.Datacenters {
+		// Determine which failover definitions apply.
-					// Rewrite the target as per the failover policy.
+		var failoverTargets []*structs.DiscoveryTarget
-					failoverTarget := c.rewriteTarget(
+		if len(failover.Datacenters) > 0 {
-						target,
+			opts := failover.ToDiscoveryTargetOpts()
-						failover.Service,
+			for _, dc := range failover.Datacenters {
-						failover.ServiceSubset,
-						target.Partition,
-						failover.Namespace,
-						dc,
-					)
-					if failoverTarget.ID != target.ID { // don't failover to yourself
-						failoverTargets = append(failoverTargets, failoverTarget)
-					}
-				}
-			} else {
 				// Rewrite the target as per the failover policy.
-				failoverTarget := c.rewriteTarget(
+				opts.Datacenter = dc
-					target,
+				failoverTarget := c.rewriteTarget(target, opts)
-					failover.Service,
-					failover.ServiceSubset,
-					target.Partition,
-					failover.Namespace,
-					"",
-				)
 				if failoverTarget.ID != target.ID { // don't failover to yourself
 					failoverTargets = append(failoverTargets, failoverTarget)
 				}
 			}
-
+		} else if len(failover.Targets) > 0 {
-			// If we filtered everything out then no point in having a failover.
+			for _, t := range failover.Targets {
-			if len(failoverTargets) > 0 {
+				// Rewrite the target as per the failover policy.
-				df := &structs.DiscoveryFailover{}
+				failoverTarget := c.rewriteTarget(target, t.ToDiscoveryTargetOpts())
-				node.Resolver.Failover = df
+				if failoverTarget.ID != target.ID { // don't failover to yourself
-
+					failoverTargets = append(failoverTargets, failoverTarget)
-				// Take care of doing any redirects or configuration loading
-				// related to targets by cheating a bit and recursing into
-				// ourselves.
-				for _, target := range failoverTargets {
-					failoverResolveNode, err := c.getResolverNode(target, true)
-					if err != nil {
-						return nil, err
-					}
-					failoverTarget := failoverResolveNode.Resolver.Target
-					df.Targets = append(df.Targets, failoverTarget)
 				}
 			}
+		} else {
+			// Rewrite the target as per the failover policy.
+			failoverTarget := c.rewriteTarget(target, failover.ToDiscoveryTargetOpts())
+			if failoverTarget.ID != target.ID { // don't failover to yourself
+				failoverTargets = append(failoverTargets, failoverTarget)
+			}
+		}
+
+		// If we filtered everything out then no point in having a failover.
+		if len(failoverTargets) > 0 {
+			df := &structs.DiscoveryFailover{}
+			node.Resolver.Failover = df
+
+			// Take care of doing any redirects or configuration loading
+			// related to targets by cheating a bit and recursing into
+			// ourselves.
+			for _, target := range failoverTargets {
+				failoverResolveNode, err := c.getResolverNode(target, true)
+				if err != nil {
+					return nil, err
+				}
+				failoverTarget := failoverResolveNode.Resolver.Target
+				df.Targets = append(df.Targets, failoverTarget)
+			}
 		}
 	}
 

agent/consul/discoverychain/compile_test.go

@@ -39,6 +39,7 @@ func TestCompile(t *testing.T) {
 		"service redirect":                                 testcase_ServiceRedirect(),
 		"service and subset redirect":                      testcase_ServiceAndSubsetRedirect(),
 		"datacenter redirect":                              testcase_DatacenterRedirect(),
+		"redirect to cluster peer":                         testcase_PeerRedirect(),
 		"datacenter redirect with mesh gateways":           testcase_DatacenterRedirect_WithMeshGateways(),
 		"service failover":                                 testcase_ServiceFailover(),
 		"service failover through redirect":                testcase_ServiceFailoverThroughRedirect(),
@@ -46,6 +47,7 @@ func TestCompile(t *testing.T) {
 		"service and subset failover":                      testcase_ServiceAndSubsetFailover(),
 		"datacenter failover":                              testcase_DatacenterFailover(),
 		"datacenter failover with mesh gateways":           testcase_DatacenterFailover_WithMeshGateways(),
+		"target failover":                                  testcase_Failover_Targets(),
 		"noop split to resolver with default subset":       testcase_NoopSplit_WithDefaultSubset(),
 		"resolver with default subset":                     testcase_Resolve_WithDefaultSubset(),
 		"default resolver with external sni":               testcase_DefaultResolver_ExternalSNI(),
@@ -182,7 +184,7 @@ func testcase_JustRouterWithDefaults() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -244,7 +246,7 @@ func testcase_JustRouterWithNoDestination() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -294,7 +296,7 @@ func testcase_RouterWithDefaults_NoSplit_WithResolver() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			"main.default.default.dc1": targetWithConnectTimeout(
-				newTarget("main", "", "default", "default", "dc1", nil),
+				newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 				33*time.Second,
 			),
 		},
@@ -361,7 +363,7 @@ func testcase_RouterWithDefaults_WithNoopSplit_DefaultResolver() compileTestCase
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -426,7 +428,10 @@ func testcase_NoopSplit_DefaultResolver_ProtocolFromProxyDefaults() compileTestC
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc1",
+			}, nil),
 		},
 	}
 
@@ -498,7 +503,7 @@ func testcase_RouterWithDefaults_WithNoopSplit_WithResolver() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			"main.default.default.dc1": targetWithConnectTimeout(
-				newTarget("main", "", "default", "default", "dc1", nil),
+				newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 				33*time.Second,
 			),
 		},
@@ -584,8 +589,11 @@ func testcase_RouteBypassesSplit() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"bypass.other.default.default.dc1": newTarget("other", "bypass", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"bypass.other.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "other",
+				ServiceSubset: "bypass",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == bypass",
 				}
@@ -638,7 +646,7 @@ func testcase_NoopSplit_DefaultResolver() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -694,7 +702,7 @@ func testcase_NoopSplit_WithResolver() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			"main.default.default.dc1": targetWithConnectTimeout(
-				newTarget("main", "", "default", "default", "dc1", nil),
+				newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 				33*time.Second,
 			),
 		},
@@ -776,12 +784,19 @@ func testcase_SubsetSplit() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"v2.main.default.default.dc1": newTarget("main", "v2", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+
+			"v2.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 2",
 				}
 			}),
-			"v1.main.default.default.dc1": newTarget("main", "v1", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v1.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v1",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 1",
 				}
@@ -855,8 +870,8 @@ func testcase_ServiceSplit() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"foo.default.default.dc1": newTarget("foo", "", "default", "default", "dc1", nil),
+			"foo.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "foo"}, nil),
-			"bar.default.default.dc1": newTarget("bar", "", "default", "default", "dc1", nil),
+			"bar.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "bar"}, nil),
 		},
 	}
 
@@ -935,7 +950,10 @@ func testcase_SplitBypassesSplit() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"bypassed.next.default.default.dc1": newTarget("next", "bypassed", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"bypassed.next.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "next",
+				ServiceSubset: "bypassed",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == bypass",
 				}
@@ -973,7 +991,7 @@ func testcase_ServiceRedirect() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"other.default.default.dc1": newTarget("other", "", "default", "default", "dc1", nil),
+			"other.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "other"}, nil),
 		},
 	}
 
@@ -1019,7 +1037,10 @@ func testcase_ServiceAndSubsetRedirect() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"v2.other.default.default.dc1": newTarget("other", "v2", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v2.other.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "other",
+				ServiceSubset: "v2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 2",
 				}
@@ -1055,7 +1076,51 @@ func testcase_DatacenterRedirect() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc9": newTarget("main", "", "default", "default", "dc9", nil),
+			"main.default.default.dc9": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc9",
+			}, nil),
+		},
+	}
+	return compileTestCase{entries: entries, expect: expect}
+}
+
+func testcase_PeerRedirect() compileTestCase {
+	entries := newEntries()
+	entries.AddResolvers(
+		&structs.ServiceResolverConfigEntry{
+			Kind: "service-resolver",
+			Name: "main",
+			Redirect: &structs.ServiceResolverRedirect{
+				Service: "other",
+				Peer:    "cluster-01",
+			},
+		},
+	)
+
+	expect := &structs.CompiledDiscoveryChain{
+		Protocol:  "tcp",
+		StartNode: "resolver:other.default.default.external.cluster-01",
+		Nodes: map[string]*structs.DiscoveryGraphNode{
+			"resolver:other.default.default.external.cluster-01": {
+				Type: structs.DiscoveryGraphNodeTypeResolver,
+				Name: "other.default.default.external.cluster-01",
+				Resolver: &structs.DiscoveryResolver{
+					Default:        true,
+					ConnectTimeout: 5 * time.Second,
+					Target:         "other.default.default.external.cluster-01",
+				},
+			},
+		},
+		Targets: map[string]*structs.DiscoveryTarget{
+			"other.default.default.external.cluster-01": newTarget(structs.DiscoveryTargetOpts{
+				Service: "other",
+				Peer:    "cluster-01",
+			}, func(t *structs.DiscoveryTarget) {
+				t.SNI = ""
+				t.Name = ""
+				t.Datacenter = ""
+			}),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1095,7 +1160,10 @@ func testcase_DatacenterRedirect_WithMeshGateways() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc9": newTarget("main", "", "default", "default", "dc9", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc9": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc9",
+			}, func(t *structs.DiscoveryTarget) {
 				t.MeshGateway = structs.MeshGatewayConfig{
 					Mode: structs.MeshGatewayModeRemote,
 				}
@@ -1134,8 +1202,8 @@ func testcase_ServiceFailover() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1":   newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1":   newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"backup.default.default.dc1": newTarget("backup", "", "default", "default", "dc1", nil),
+			"backup.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "backup"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1177,8 +1245,8 @@ func testcase_ServiceFailoverThroughRedirect() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1":   newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1":   newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"actual.default.default.dc1": newTarget("actual", "", "default", "default", "dc1", nil),
+			"actual.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "actual"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1220,8 +1288,8 @@ func testcase_Resolver_CircularFailover() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1":   newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1":   newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"backup.default.default.dc1": newTarget("backup", "", "default", "default", "dc1", nil),
+			"backup.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "backup"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1261,8 +1329,11 @@ func testcase_ServiceAndSubsetFailover() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"backup.main.default.default.dc1": newTarget("main", "backup", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"backup.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "backup",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == backup",
 				}
@@ -1301,9 +1372,15 @@ func testcase_DatacenterFailover() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
-			"main.default.default.dc2": newTarget("main", "", "default", "default", "dc2", nil),
+			"main.default.default.dc2": newTarget(structs.DiscoveryTargetOpts{
-			"main.default.default.dc4": newTarget("main", "", "default", "default", "dc4", nil),
+				Service:    "main",
+				Datacenter: "dc2",
+			}, nil),
+			"main.default.default.dc4": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc4",
+			}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1350,17 +1427,105 @@ func testcase_DatacenterFailover_WithMeshGateways() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, func(t *structs.DiscoveryTarget) {
 				t.MeshGateway = structs.MeshGatewayConfig{
 					Mode: structs.MeshGatewayModeRemote,
 				}
 			}),
-			"main.default.default.dc2": newTarget("main", "", "default", "default", "dc2", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc2": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.MeshGateway = structs.MeshGatewayConfig{
 					Mode: structs.MeshGatewayModeRemote,
 				}
 			}),
-			"main.default.default.dc4": newTarget("main", "", "default", "default", "dc4", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc4": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc4",
+			}, func(t *structs.DiscoveryTarget) {
+				t.MeshGateway = structs.MeshGatewayConfig{
+					Mode: structs.MeshGatewayModeRemote,
+				}
+			}),
+		},
+	}
+	return compileTestCase{entries: entries, expect: expect}
+}
+
+func testcase_Failover_Targets() compileTestCase {
+	entries := newEntries()
+
+	entries.AddProxyDefaults(&structs.ProxyConfigEntry{
+		Kind: structs.ProxyDefaults,
+		Name: structs.ProxyConfigGlobal,
+		MeshGateway: structs.MeshGatewayConfig{
+			Mode: structs.MeshGatewayModeRemote,
+		},
+	})
+
+	entries.AddResolvers(
+		&structs.ServiceResolverConfigEntry{
+			Kind: "service-resolver",
+			Name: "main",
+			Failover: map[string]structs.ServiceResolverFailover{
+				"*": {
+					Targets: []structs.ServiceResolverFailoverTarget{
+						{Datacenter: "dc3"},
+						{Service: "new-main"},
+						{Peer: "cluster-01"},
+					},
+				},
+			},
+		},
+	)
+
+	expect := &structs.CompiledDiscoveryChain{
+		Protocol:  "tcp",
+		StartNode: "resolver:main.default.default.dc1",
+		Nodes: map[string]*structs.DiscoveryGraphNode{
+			"resolver:main.default.default.dc1": {
+				Type: structs.DiscoveryGraphNodeTypeResolver,
+				Name: "main.default.default.dc1",
+				Resolver: &structs.DiscoveryResolver{
+					ConnectTimeout: 5 * time.Second,
+					Target:         "main.default.default.dc1",
+					Failover: &structs.DiscoveryFailover{
+						Targets: []string{
+							"main.default.default.dc3",
+							"new-main.default.default.dc1",
+							"main.default.default.external.cluster-01",
+						},
+					},
+				},
+			},
+		},
+		Targets: map[string]*structs.DiscoveryTarget{
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, func(t *structs.DiscoveryTarget) {
+				t.MeshGateway = structs.MeshGatewayConfig{
+					Mode: structs.MeshGatewayModeRemote,
+				}
+			}),
+			"main.default.default.dc3": newTarget(structs.DiscoveryTargetOpts{
+				Service:    "main",
+				Datacenter: "dc3",
+			}, func(t *structs.DiscoveryTarget) {
+				t.MeshGateway = structs.MeshGatewayConfig{
+					Mode: structs.MeshGatewayModeRemote,
+				}
+			}),
+			"new-main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "new-main"}, func(t *structs.DiscoveryTarget) {
+				t.MeshGateway = structs.MeshGatewayConfig{
+					Mode: structs.MeshGatewayModeRemote,
+				}
+			}),
+			"main.default.default.external.cluster-01": newTarget(structs.DiscoveryTargetOpts{
+				Service: "main",
+				Peer:    "cluster-01",
+			}, func(t *structs.DiscoveryTarget) {
+				t.SNI = ""
+				t.Name = ""
+				t.Datacenter = ""
 				t.MeshGateway = structs.MeshGatewayConfig{
 					Mode: structs.MeshGatewayModeRemote,
 				}
@@ -1422,7 +1587,10 @@ func testcase_NoopSplit_WithDefaultSubset() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"v2.main.default.default.dc1": newTarget("main", "v2", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v2.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 2",
 				}
@@ -1452,7 +1620,7 @@ func testcase_DefaultResolver() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			// TODO-TARGET
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect}
@@ -1488,7 +1656,7 @@ func testcase_DefaultResolver_WithProxyDefaults() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, func(t *structs.DiscoveryTarget) {
 				t.MeshGateway = structs.MeshGatewayConfig{
 					Mode: structs.MeshGatewayModeRemote,
 				}
@@ -1530,7 +1698,7 @@ func testcase_ServiceMetaProjection() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -1588,7 +1756,7 @@ func testcase_ServiceMetaProjectionWithRedirect() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"other.default.default.dc1": newTarget("other", "", "default", "default", "dc1", nil),
+			"other.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "other"}, nil),
 		},
 	}
 
@@ -1623,7 +1791,7 @@ func testcase_RedirectToDefaultResolverIsNotDefaultChain() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"other.default.default.dc1": newTarget("other", "", "default", "default", "dc1", nil),
+			"other.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "other"}, nil),
 		},
 	}
 
@@ -1658,7 +1826,10 @@ func testcase_Resolve_WithDefaultSubset() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"v2.main.default.default.dc1": newTarget("main", "v2", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v2.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 2",
 				}
@@ -1692,7 +1863,7 @@ func testcase_DefaultResolver_ExternalSNI() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, func(t *structs.DiscoveryTarget) {
 				t.SNI = "main.some.other.service.mesh"
 				t.External = true
 			}),
@@ -1857,11 +2028,17 @@ func testcase_MultiDatacenterCanary() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			"main.default.default.dc2": targetWithConnectTimeout(
-				newTarget("main", "", "default", "default", "dc2", nil),
+				newTarget(structs.DiscoveryTargetOpts{
+					Service:    "main",
+					Datacenter: "dc2",
+				}, nil),
 				33*time.Second,
 			),
 			"main.default.default.dc3": targetWithConnectTimeout(
-				newTarget("main", "", "default", "default", "dc3", nil),
+				newTarget(structs.DiscoveryTargetOpts{
+					Service:    "main",
+					Datacenter: "dc3",
+				}, nil),
 				33*time.Second,
 			),
 		},
@@ -2155,27 +2332,42 @@ func testcase_AllBellsAndWhistles() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"prod.redirected.default.default.dc1": newTarget("redirected", "prod", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"prod.redirected.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "redirected",
+				ServiceSubset: "prod",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "ServiceMeta.env == prod",
 				}
 			}),
-			"v1.main.default.default.dc1": newTarget("main", "v1", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v1.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v1",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 1",
 				}
 			}),
-			"v2.main.default.default.dc1": newTarget("main", "v2", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v2.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v2",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 2",
 				}
 			}),
-			"v3.main.default.default.dc1": newTarget("main", "v3", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"v3.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "v3",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{
 					Filter: "Service.Meta.version == 3",
 				}
 			}),
-			"default-subset.main.default.default.dc1": newTarget("main", "default-subset", "default", "default", "dc1", func(t *structs.DiscoveryTarget) {
+			"default-subset.main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{
+				Service:       "main",
+				ServiceSubset: "default-subset",
+			}, func(t *structs.DiscoveryTarget) {
 				t.Subset = structs.ServiceResolverSubset{OnlyPassing: true}
 			}),
 		},
@@ -2379,7 +2571,7 @@ func testcase_ResolverProtocolOverride() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			// TODO-TARGET
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect,
@@ -2413,7 +2605,7 @@ func testcase_ResolverProtocolOverrideIgnored() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			// TODO-TARGET
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect,
@@ -2451,7 +2643,7 @@ func testcase_RouterIgnored_ResolverProtocolOverride() compileTestCase {
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
 			// TODO-TARGET
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 	return compileTestCase{entries: entries, expect: expect,
@@ -2685,9 +2877,9 @@ func testcase_LBSplitterAndResolver() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"foo.default.default.dc1": newTarget("foo", "", "default", "default", "dc1", nil),
+			"foo.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "foo"}, nil),
-			"bar.default.default.dc1": newTarget("bar", "", "default", "default", "dc1", nil),
+			"bar.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "bar"}, nil),
-			"baz.default.default.dc1": newTarget("baz", "", "default", "default", "dc1", nil),
+			"baz.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "baz"}, nil),
 		},
 	}
 
@@ -2743,7 +2935,7 @@ func testcase_LBResolver() compileTestCase {
 			},
 		},
 		Targets: map[string]*structs.DiscoveryTarget{
-			"main.default.default.dc1": newTarget("main", "", "default", "default", "dc1", nil),
+			"main.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "main"}, nil),
 		},
 	}
 
@@ -2791,8 +2983,17 @@ func newEntries() *configentry.DiscoveryChainSet {
 	}
 }
 
-func newTarget(service, serviceSubset, namespace, partition, datacenter string, modFn func(t *structs.DiscoveryTarget)) *structs.DiscoveryTarget {
+func newTarget(opts structs.DiscoveryTargetOpts, modFn func(t *structs.DiscoveryTarget)) *structs.DiscoveryTarget {
-	t := structs.NewDiscoveryTarget(service, serviceSubset, namespace, partition, datacenter)
+	if opts.Namespace == "" {
+		opts.Namespace = "default"
+	}
+	if opts.Partition == "" {
+		opts.Partition = "default"
+	}
+	if opts.Datacenter == "" {
+		opts.Datacenter = "dc1"
+	}
+	t := structs.NewDiscoveryTarget(opts)
 	t.SNI = connect.TargetSNI(t, "trustdomain.consul")
 	t.Name = t.SNI
 	t.ConnectTimeout = 5 * time.Second // default

agent/consul/internal_endpoint.go

@@ -153,64 +153,87 @@ func (m *Internal) ServiceDump(args *structs.ServiceDumpRequest, reply *structs.
 		&args.QueryOptions,
 		&reply.QueryMeta,
 		func(ws memdb.WatchSet, state *state.Store) error {
-			// we don't support calling this endpoint for a specific peer
-			if args.PeerName != "" {
-				return fmt.Errorf("this endpoint does not support specifying a peer: %q", args.PeerName)
-			}
-
 			// this maxIndex will be the max of the ServiceDump calls and the PeeringList call
 			var maxIndex uint64
 
-			// get a local dump for services
+			// If PeerName is not empty, we return only the imported services from that peer
-			index, nodes, err := state.ServiceDump(ws, args.ServiceKind, args.UseServiceKind, &args.EnterpriseMeta, structs.DefaultPeerKeyword)
+			if args.PeerName != "" {
-			if err != nil {
+				// get a local dump for services
-				return fmt.Errorf("could not get a service dump for local nodes: %w", err)
+				index, nodes, err := state.ServiceDump(ws,
-			}
+					args.ServiceKind,
-
+					args.UseServiceKind,
-			if index > maxIndex {
+					// Note we fetch imported services with wildcard namespace because imported services' namespaces
-				maxIndex = index
+					// are in a different locality; regardless of our local namespace, we return all imported services
-			}
+					// of the local partition.
-			reply.Nodes = nodes
+					args.EnterpriseMeta.WithWildcardNamespace(),
-
+					args.PeerName)
-			// get a list of all peerings
-			index, listedPeerings, err := state.PeeringList(ws, args.EnterpriseMeta)
-			if err != nil {
-				return fmt.Errorf("could not list peers for service dump %w", err)
-			}
-
-			if index > maxIndex {
-				maxIndex = index
-			}
-
-			for _, p := range listedPeerings {
-				index, importedNodes, err := state.ServiceDump(ws, args.ServiceKind, args.UseServiceKind, &args.EnterpriseMeta, p.Name)
 				if err != nil {
-					return fmt.Errorf("could not get a service dump for peer %q: %w", p.Name, err)
+					return fmt.Errorf("could not get a service dump for peer %q: %w", args.PeerName, err)
 				}
 
 				if index > maxIndex {
 					maxIndex = index
 				}
-				reply.ImportedNodes = append(reply.ImportedNodes, importedNodes...)
+				reply.Index = maxIndex
-			}
+				reply.ImportedNodes = nodes
 
-			// Get, store, and filter gateway services
+			} else {
-			idx, gatewayServices, err := state.DumpGatewayServices(ws)
+				// otherwise return both local and all imported services
-			if err != nil {
-				return err
-			}
-			reply.Gateways = gatewayServices
 
-			if idx > maxIndex {
+				// get a local dump for services
-				maxIndex = idx
+				index, nodes, err := state.ServiceDump(ws, args.ServiceKind, args.UseServiceKind, &args.EnterpriseMeta, structs.DefaultPeerKeyword)
-			}
+				if err != nil {
-			reply.Index = maxIndex
+					return fmt.Errorf("could not get a service dump for local nodes: %w", err)
+				}
 
-			raw, err := filter.Execute(reply.Nodes)
+				if index > maxIndex {
-			if err != nil {
+					maxIndex = index
-				return fmt.Errorf("could not filter local service dump: %w", err)
+				}
+				reply.Nodes = nodes
+
+				// get a list of all peerings
+				index, listedPeerings, err := state.PeeringList(ws, args.EnterpriseMeta)
+				if err != nil {
+					return fmt.Errorf("could not list peers for service dump %w", err)
+				}
+
+				if index > maxIndex {
+					maxIndex = index
+				}
+
+				for _, p := range listedPeerings {
+					// Note we fetch imported services with wildcard namespace because imported services' namespaces
+					// are in a different locality; regardless of our local namespace, we return all imported services
+					// of the local partition.
+					index, importedNodes, err := state.ServiceDump(ws, args.ServiceKind, args.UseServiceKind, args.EnterpriseMeta.WithWildcardNamespace(), p.Name)
+					if err != nil {
+						return fmt.Errorf("could not get a service dump for peer %q: %w", p.Name, err)
+					}
+
+					if index > maxIndex {
+						maxIndex = index
+					}
+					reply.ImportedNodes = append(reply.ImportedNodes, importedNodes...)
+				}
+
+				// Get, store, and filter gateway services
+				idx, gatewayServices, err := state.DumpGatewayServices(ws)
+				if err != nil {
+					return err
+				}
+				reply.Gateways = gatewayServices
+
+				if idx > maxIndex {
+					maxIndex = idx
+				}
+				reply.Index = maxIndex
+
+				raw, err := filter.Execute(reply.Nodes)
+				if err != nil {
+					return fmt.Errorf("could not filter local service dump: %w", err)
+				}
+				reply.Nodes = raw.(structs.CheckServiceNodes)
 			}
-			reply.Nodes = raw.(structs.CheckServiceNodes)
 
 			importedRaw, err := filter.Execute(reply.ImportedNodes)
 			if err != nil {

agent/consul/leader_connect_ca.go

@@ -1098,11 +1098,36 @@ func setLeafSigningCert(caRoot *structs.CARoot, pem string) error {
 		return fmt.Errorf("error parsing leaf signing cert: %w", err)
 	}
 
+	if err := pruneExpiredIntermediates(caRoot); err != nil {
+		return err
+	}
+
 	caRoot.IntermediateCerts = append(caRoot.IntermediateCerts, pem)
 	caRoot.SigningKeyID = connect.EncodeSigningKeyID(cert.SubjectKeyId)
 	return nil
 }
 
+// pruneExpiredIntermediates removes expired intermediate certificates
+// from the given CARoot.
+func pruneExpiredIntermediates(caRoot *structs.CARoot) error {
+	var newIntermediates []string
+	now := time.Now()
+	for _, intermediatePEM := range caRoot.IntermediateCerts {
+		cert, err := connect.ParseCert(intermediatePEM)
+		if err != nil {
+			return fmt.Errorf("error parsing leaf signing cert: %w", err)
+		}
+
+		// Only keep the intermediate cert if it's still valid.
+		if cert.NotAfter.After(now) {
+			newIntermediates = append(newIntermediates, intermediatePEM)
+		}
+	}
+
+	caRoot.IntermediateCerts = newIntermediates
+	return nil
+}
+
 // runRenewIntermediate periodically attempts to renew the intermediate cert.
 func (c *CAManager) runRenewIntermediate(ctx context.Context) error {
 	isPrimary := c.serverConf.Datacenter == c.serverConf.PrimaryDatacenter

agent/consul/leader_connect_ca_test.go

@@ -435,7 +435,6 @@ func TestCAManager_SignCertificate_WithExpiredCert(t *testing.T) {
 		errorMsg              string
 	}{
 		{"intermediate valid", time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, 2), time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, 2), false, ""},
-		{"intermediate expired", time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, 2), time.Now().AddDate(-2, 0, 0), time.Now().AddDate(0, 0, -1), true, "intermediate expired: certificate expired, expiration date"},
 		{"root expired", time.Now().AddDate(-2, 0, 0), time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, 2), true, "root expired: certificate expired, expiration date"},
 		// a cert that is not yet valid is ok, assume it will be valid soon enough
 		{"intermediate in the future", time.Now().AddDate(0, 0, -1), time.Now().AddDate(0, 0, 2), time.Now().AddDate(0, 0, 1), time.Now().AddDate(0, 0, 2), false, ""},

agent/consul/leader_connect_test.go

@@ -401,6 +401,18 @@ func TestCAManager_RenewIntermediate_Vault_Primary(t *testing.T) {
 	err = msgpackrpc.CallWithCodec(codec, "ConnectCA.Sign", &req, &cert)
 	require.NoError(t, err)
 	verifyLeafCert(t, activeRoot, cert.CertPEM)
+
+	// Wait for the primary's old intermediate to be pruned after expiring.
+	oldIntermediate := activeRoot.IntermediateCerts[0]
+	retry.Run(t, func(r *retry.R) {
+		store := s1.caManager.delegate.State()
+		_, storedRoot, err := store.CARootActive(nil)
+		r.Check(err)
+
+		if storedRoot.IntermediateCerts[0] == oldIntermediate {
+			r.Fatal("old intermediate should be gone")
+		}
+	})
 }
 
 func patchIntermediateCertRenewInterval(t *testing.T) {
@@ -516,6 +528,18 @@ func TestCAManager_RenewIntermediate_Secondary(t *testing.T) {
 	err = msgpackrpc.CallWithCodec(codec, "ConnectCA.Sign", &req, &cert)
 	require.NoError(t, err)
 	verifyLeafCert(t, activeRoot, cert.CertPEM)
+
+	// Wait for dc2's old intermediate to be pruned after expiring.
+	oldIntermediate := activeRoot.IntermediateCerts[0]
+	retry.Run(t, func(r *retry.R) {
+		store := s2.caManager.delegate.State()
+		_, storedRoot, err := store.CARootActive(nil)
+		r.Check(err)
+
+		if storedRoot.IntermediateCerts[0] == oldIntermediate {
+			r.Fatal("old intermediate should be gone")
+		}
+	})
 }
 
 func TestConnectCA_ConfigurationSet_RootRotation_Secondary(t *testing.T) {

agent/consul/leader_peering.go

@@ -31,11 +31,18 @@ import (
 )
 
 var leaderExportedServicesCountKey = []string{"consul", "peering", "exported_services"}
+var leaderHealthyPeeringKey = []string{"consul", "peering", "healthy"}
 var LeaderPeeringMetrics = []prometheus.GaugeDefinition{
 	{
 		Name: leaderExportedServicesCountKey,
 		Help: "A gauge that tracks how many services are exported for the peering. " +
-			"The labels are \"peering\" and, for enterprise, \"partition\". " +
+			"The labels are \"peer_name\", \"peer_id\" and, for enterprise, \"partition\". " +
+			"We emit this metric every 9 seconds",
+	},
+	{
+		Name: leaderHealthyPeeringKey,
+		Help: "A gauge that tracks how if a peering is healthy (1) or not (0). " +
+			"The labels are \"peer_name\", \"peer_id\" and, for enterprise, \"partition\". " +
 			"We emit this metric every 9 seconds",
 	},
 }
@@ -85,13 +92,6 @@ func (s *Server) emitPeeringMetricsOnce(logger hclog.Logger, metricsImpl *metric
 	}
 
 	for _, peer := range peers {
-		status, found := s.peerStreamServer.StreamStatus(peer.ID)
-		if !found {
-			logger.Trace("did not find status for", "peer_name", peer.Name)
-			continue
-		}
-
-		esc := status.GetExportedServicesCount()
 		part := peer.Partition
 		labels := []metrics.Label{
 			{Name: "peer_name", Value: peer.Name},
@@ -101,7 +101,25 @@ func (s *Server) emitPeeringMetricsOnce(logger hclog.Logger, metricsImpl *metric
 			labels = append(labels, metrics.Label{Name: "partition", Value: part})
 		}
 
-		metricsImpl.SetGaugeWithLabels(leaderExportedServicesCountKey, float32(esc), labels)
+		status, found := s.peerStreamServer.StreamStatus(peer.ID)
+		if found {
+			// exported services count metric
+			esc := status.GetExportedServicesCount()
+			metricsImpl.SetGaugeWithLabels(leaderExportedServicesCountKey, float32(esc), labels)
+		}
+
+		// peering health metric
+		if status.NeverConnected {
+			metricsImpl.SetGaugeWithLabels(leaderHealthyPeeringKey, float32(math.NaN()), labels)
+		} else {
+			healthy := s.peerStreamServer.Tracker.IsHealthy(status)
+			healthyInt := 0
+			if healthy {
+				healthyInt = 1
+			}
+
+			metricsImpl.SetGaugeWithLabels(leaderHealthyPeeringKey, float32(healthyInt), labels)
+		}
 	}
 
 	return nil
@@ -277,13 +295,6 @@ func (s *Server) establishStream(ctx context.Context, logger hclog.Logger, ws me
 		return fmt.Errorf("failed to build TLS dial option from peering: %w", err)
 	}
 
-	// Create a ring buffer to cycle through peer addresses in the retry loop below.
-	buffer := ring.New(len(peer.PeerServerAddresses))
-	for _, addr := range peer.PeerServerAddresses {
-		buffer.Value = addr
-		buffer = buffer.Next()
-	}
-
 	secret, err := s.fsm.State().PeeringSecretsRead(ws, peer.ID)
 	if err != nil {
 		return fmt.Errorf("failed to read secret for peering: %w", err)
@@ -294,27 +305,26 @@ func (s *Server) establishStream(ctx context.Context, logger hclog.Logger, ws me
 
 	logger.Trace("establishing stream to peer")
 
-	retryCtx, cancel := context.WithCancel(ctx)
+	streamStatus, err := s.peerStreamServer.Tracker.Register(peer.ID)
-	cancelFns[peer.ID] = cancel
-
-	streamStatus, err := s.peerStreamTracker.Register(peer.ID)
 	if err != nil {
 		return fmt.Errorf("failed to register stream: %v", err)
 	}
 
+	streamCtx, cancel := context.WithCancel(ctx)
+	cancelFns[peer.ID] = cancel
+
+	// Start a goroutine to watch for updates to peer server addresses.
+	// The latest valid server address can be received from nextServerAddr.
+	nextServerAddr := make(chan string)
+	go s.watchPeerServerAddrs(streamCtx, peer, nextServerAddr)
+
 	// Establish a stream-specific retry so that retrying stream/conn errors isn't dependent on state store changes.
-	go retryLoopBackoffPeering(retryCtx, logger, func() error {
+	go retryLoopBackoffPeering(streamCtx, logger, func() error {
 		// Try a new address on each iteration by advancing the ring buffer on errors.
-		defer func() {
+		addr := <-nextServerAddr
-			buffer = buffer.Next()
-		}()
-		addr, ok := buffer.Value.(string)
-		if !ok {
-			return fmt.Errorf("peer server address type %T is not a string", buffer.Value)
-		}
 
 		logger.Trace("dialing peer", "addr", addr)
-		conn, err := grpc.DialContext(retryCtx, addr,
+		conn, err := grpc.DialContext(streamCtx, addr,
 			// TODO(peering): use a grpc.WithStatsHandler here?)
 			tlsOption,
 			// For keep alive parameters there is a larger comment in ClientConnPool.dial about that.
@@ -331,7 +341,7 @@ func (s *Server) establishStream(ctx context.Context, logger hclog.Logger, ws me
 		defer conn.Close()
 
 		client := pbpeerstream.NewPeerStreamServiceClient(conn)
-		stream, err := client.StreamResources(retryCtx)
+		stream, err := client.StreamResources(streamCtx)
 		if err != nil {
 			return err
 		}
@@ -379,6 +389,74 @@ func (s *Server) establishStream(ctx context.Context, logger hclog.Logger, ws me
 	return nil
 }
 
+// watchPeerServerAddrs sends an up-to-date peer server address to nextServerAddr.
+// It loads the server addresses into a ring buffer and cycles through them until:
+//  1. streamCtx is cancelled (peer is deleted)
+//  2. the peer is modified and the watchset fires.
+//
+// In case (2) we refetch the peering and rebuild the ring buffer.
+func (s *Server) watchPeerServerAddrs(ctx context.Context, peer *pbpeering.Peering, nextServerAddr chan<- string) {
+	defer close(nextServerAddr)
+
+	// we initialize the ring buffer with the peer passed to `establishStream`
+	// because the caller has pre-checked `peer.ShouldDial`, guaranteeing
+	// at least one server address.
+	//
+	// IMPORTANT: ringbuf must always be length > 0 or else `<-nextServerAddr` may block.
+	ringbuf := ring.New(len(peer.PeerServerAddresses))
+	for _, addr := range peer.PeerServerAddresses {
+		ringbuf.Value = addr
+		ringbuf = ringbuf.Next()
+	}
+	innerWs := memdb.NewWatchSet()
+	_, _, err := s.fsm.State().PeeringReadByID(innerWs, peer.ID)
+	if err != nil {
+		s.logger.Warn("failed to watch for changes to peer; server addresses may become stale over time.",
+			"peer_id", peer.ID,
+			"error", err)
+	}
+
+	fetchAddrs := func() error {
+		// reinstantiate innerWs to prevent it from growing indefinitely
+		innerWs = memdb.NewWatchSet()
+		_, peering, err := s.fsm.State().PeeringReadByID(innerWs, peer.ID)
+		if err != nil {
+			return fmt.Errorf("failed to fetch peer %q: %w", peer.ID, err)
+		}
+		if !peering.IsActive() {
+			return fmt.Errorf("peer %q is no longer active", peer.ID)
+		}
+		if len(peering.PeerServerAddresses) == 0 {
+			return fmt.Errorf("peer %q has no addresses to dial", peer.ID)
+		}
+
+		ringbuf = ring.New(len(peering.PeerServerAddresses))
+		for _, addr := range peering.PeerServerAddresses {
+			ringbuf.Value = addr
+			ringbuf = ringbuf.Next()
+		}
+		return nil
+	}
+
+	for {
+		select {
+		case nextServerAddr <- ringbuf.Value.(string):
+			ringbuf = ringbuf.Next()
+		case err := <-innerWs.WatchCh(ctx):
+			if err != nil {
+				// context was cancelled
+				return
+			}
+			// watch fired so we refetch the peering and rebuild the ring buffer
+			if err := fetchAddrs(); err != nil {
+				s.logger.Warn("watchset for peer was fired but failed to update server addresses",
+					"peer_id", peer.ID,
+					"error", err)
+			}
+		}
+	}
+}
+
 func (s *Server) startPeeringDeferredDeletion(ctx context.Context) {
 	s.leaderRoutineManager.Start(ctx, peeringDeletionRoutineName, s.runPeeringDeletions)
 }
@@ -391,6 +469,12 @@ func (s *Server) runPeeringDeletions(ctx context.Context) error {
 	// process. This includes deletion of the peerings themselves in addition to any peering data
 	raftLimiter := rate.NewLimiter(defaultDeletionApplyRate, int(defaultDeletionApplyRate))
 	for {
+		select {
+		case <-ctx.Done():
+			return nil
+		default:
+		}
+
 		ws := memdb.NewWatchSet()
 		state := s.fsm.State()
 		_, peerings, err := s.fsm.State().PeeringListDeleted(ws)

agent/consul/leader_peering_test.go

@@ -7,6 +7,7 @@ import (
 	"errors"
 	"fmt"
 	"io/ioutil"
+	"math"
 	"testing"
 	"time"
 
@@ -17,6 +18,7 @@ import (
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/codes"
 	grpcstatus "google.golang.org/grpc/status"
+	"google.golang.org/protobuf/proto"
 
 	"github.com/hashicorp/consul/acl"
 	"github.com/hashicorp/consul/agent/consul/state"
@@ -24,6 +26,7 @@ import (
 	"github.com/hashicorp/consul/api"
 	"github.com/hashicorp/consul/proto/pbpeering"
 	"github.com/hashicorp/consul/sdk/freeport"
+	"github.com/hashicorp/consul/sdk/testutil"
 	"github.com/hashicorp/consul/sdk/testutil/retry"
 	"github.com/hashicorp/consul/testrpc"
 	"github.com/hashicorp/consul/types"
@@ -37,6 +40,7 @@ func TestLeader_PeeringSync_Lifecycle_ClientDeletion(t *testing.T) {
 		testLeader_PeeringSync_Lifecycle_ClientDeletion(t, true)
 	})
 }
+
 func testLeader_PeeringSync_Lifecycle_ClientDeletion(t *testing.T, enableTLS bool) {
 	if testing.Short() {
 		t.Skip("too slow for testing.Short")
@@ -134,9 +138,11 @@ func testLeader_PeeringSync_Lifecycle_ClientDeletion(t *testing.T, enableTLS boo
 
 	// Delete the peering to trigger the termination sequence.
 	deleted := &pbpeering.Peering{
-		ID:        p.Peering.ID,
+		ID:                  p.Peering.ID,
-		Name:      "my-peer-acceptor",
+		Name:                "my-peer-acceptor",
-		DeletedAt: structs.TimeToProto(time.Now()),
+		State:               pbpeering.PeeringState_DELETING,
+		PeerServerAddresses: p.Peering.PeerServerAddresses,
+		DeletedAt:           structs.TimeToProto(time.Now()),
 	}
 	require.NoError(t, dialer.fsm.State().PeeringWrite(2000, &pbpeering.PeeringWriteRequest{Peering: deleted}))
 	dialer.logger.Trace("deleted peering for my-peer-acceptor")
@@ -259,6 +265,7 @@ func testLeader_PeeringSync_Lifecycle_AcceptorDeletion(t *testing.T, enableTLS b
 	deleted := &pbpeering.Peering{
 		ID:        p.Peering.PeerID,
 		Name:      "my-peer-dialer",
+		State:     pbpeering.PeeringState_DELETING,
 		DeletedAt: structs.TimeToProto(time.Now()),
 	}
 
@@ -428,6 +435,7 @@ func TestLeader_Peering_DeferredDeletion(t *testing.T) {
 		Peering: &pbpeering.Peering{
 			ID:        peerID,
 			Name:      peerName,
+			State:     pbpeering.PeeringState_DELETING,
 			DeletedAt: structs.TimeToProto(time.Now()),
 		},
 	}))
@@ -974,6 +982,7 @@ func TestLeader_PeeringMetrics_emitPeeringMetrics(t *testing.T) {
 	var (
 		s2PeerID1          = generateUUID()
 		s2PeerID2          = generateUUID()
+		s2PeerID3          = generateUUID()
 		testContextTimeout = 60 * time.Second
 		lastIdx            = uint64(0)
 	)
@@ -1063,6 +1072,24 @@ func TestLeader_PeeringMetrics_emitPeeringMetrics(t *testing.T) {
 		// mimic tracking exported services
 		mst2.TrackExportedService(structs.ServiceName{Name: "d-service"})
 		mst2.TrackExportedService(structs.ServiceName{Name: "e-service"})
+
+		// pretend that the hearbeat happened
+		mst2.TrackRecvHeartbeat()
+	}
+
+	// Simulate a peering that never connects
+	{
+		p3 := &pbpeering.Peering{
+			ID:                  s2PeerID3,
+			Name:                "my-peer-s4",
+			PeerID:              token.PeerID, // doesn't much matter what these values are
+			PeerCAPems:          token.CA,
+			PeerServerName:      token.ServerName,
+			PeerServerAddresses: token.ServerAddresses,
+		}
+		require.True(t, p3.ShouldDial())
+		lastIdx++
+		require.NoError(t, s2.fsm.State().PeeringWrite(lastIdx, &pbpeering.PeeringWriteRequest{Peering: p3}))
 	}
 
 	// set up a metrics sink
@@ -1092,6 +1119,18 @@ func TestLeader_PeeringMetrics_emitPeeringMetrics(t *testing.T) {
 		require.True(r, ok, fmt.Sprintf("did not find the key %q", keyMetric2))
 
 		require.Equal(r, float32(2), metric2.Value) // for d, e services
+
+		keyHealthyMetric2 := fmt.Sprintf("us-west.consul.peering.healthy;peer_name=my-peer-s3;peer_id=%s", s2PeerID2)
+		healthyMetric2, ok := intv.Gauges[keyHealthyMetric2]
+		require.True(r, ok, fmt.Sprintf("did not find the key %q", keyHealthyMetric2))
+
+		require.Equal(r, float32(1), healthyMetric2.Value)
+
+		keyHealthyMetric3 := fmt.Sprintf("us-west.consul.peering.healthy;peer_name=my-peer-s4;peer_id=%s", s2PeerID3)
+		healthyMetric3, ok := intv.Gauges[keyHealthyMetric3]
+		require.True(r, ok, fmt.Sprintf("did not find the key %q", keyHealthyMetric3))
+
+		require.True(r, math.IsNaN(float64(healthyMetric3.Value)))
 	})
 }
 
@@ -1131,6 +1170,7 @@ func TestLeader_Peering_NoDeletionWhenPeeringDisabled(t *testing.T) {
 		Peering: &pbpeering.Peering{
 			ID:        peerID,
 			Name:      peerName,
+			State:     pbpeering.PeeringState_DELETING,
 			DeletedAt: structs.TimeToProto(time.Now()),
 		},
 	}))
@@ -1182,7 +1222,7 @@ func TestLeader_Peering_NoEstablishmentWhenPeeringDisabled(t *testing.T) {
 	}))
 
 	require.Never(t, func() bool {
-		_, found := s1.peerStreamTracker.StreamStatus(peerID)
+		_, found := s1.peerStreamServer.StreamStatus(peerID)
 		return found
 	}, 7*time.Second, 1*time.Second, "peering should not have been established")
 }
@@ -1343,3 +1383,138 @@ func Test_isFailedPreconditionErr(t *testing.T) {
 	werr := fmt.Errorf("wrapped: %w", err)
 	assert.True(t, isFailedPreconditionErr(werr))
 }
+
+func Test_Leader_PeeringSync_ServerAddressUpdates(t *testing.T) {
+	if testing.Short() {
+		t.Skip("too slow for testing.Short")
+	}
+
+	// We want 1s retries for this test
+	orig := maxRetryBackoff
+	maxRetryBackoff = 1
+	t.Cleanup(func() { maxRetryBackoff = orig })
+
+	_, acceptor := testServerWithConfig(t, func(c *Config) {
+		c.NodeName = "acceptor"
+		c.Datacenter = "dc1"
+		c.TLSConfig.Domain = "consul"
+	})
+	testrpc.WaitForLeader(t, acceptor.RPC, "dc1")
+
+	// Create a peering by generating a token
+	ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+	t.Cleanup(cancel)
+
+	conn, err := grpc.DialContext(ctx, acceptor.config.RPCAddr.String(),
+		grpc.WithContextDialer(newServerDialer(acceptor.config.RPCAddr.String())),
+		grpc.WithInsecure(),
+		grpc.WithBlock())
+	require.NoError(t, err)
+	defer conn.Close()
+
+	acceptorClient := pbpeering.NewPeeringServiceClient(conn)
+
+	req := pbpeering.GenerateTokenRequest{
+		PeerName: "my-peer-dialer",
+	}
+	resp, err := acceptorClient.GenerateToken(ctx, &req)
+	require.NoError(t, err)
+
+	// Bring up dialer and establish a peering with acceptor's token so that it attempts to dial.
+	_, dialer := testServerWithConfig(t, func(c *Config) {
+		c.NodeName = "dialer"
+		c.Datacenter = "dc2"
+		c.PrimaryDatacenter = "dc2"
+	})
+	testrpc.WaitForLeader(t, dialer.RPC, "dc2")
+
+	// Create a peering at dialer by establishing a peering with acceptor's token
+	ctx, cancel = context.WithTimeout(context.Background(), 3*time.Second)
+	t.Cleanup(cancel)
+
+	conn, err = grpc.DialContext(ctx, dialer.config.RPCAddr.String(),
+		grpc.WithContextDialer(newServerDialer(dialer.config.RPCAddr.String())),
+		grpc.WithInsecure(),
+		grpc.WithBlock())
+	require.NoError(t, err)
+	defer conn.Close()
+
+	dialerClient := pbpeering.NewPeeringServiceClient(conn)
+
+	establishReq := pbpeering.EstablishRequest{
+		PeerName:     "my-peer-acceptor",
+		PeeringToken: resp.PeeringToken,
+	}
+	_, err = dialerClient.Establish(ctx, &establishReq)
+	require.NoError(t, err)
+
+	p, err := dialerClient.PeeringRead(ctx, &pbpeering.PeeringReadRequest{Name: "my-peer-acceptor"})
+	require.NoError(t, err)
+
+	retry.Run(t, func(r *retry.R) {
+		status, found := dialer.peerStreamServer.StreamStatus(p.Peering.ID)
+		require.True(r, found)
+		require.True(r, status.Connected)
+	})
+
+	testutil.RunStep(t, "calling establish with active connection does not overwrite server addresses", func(t *testing.T) {
+		ctx, cancel = context.WithTimeout(context.Background(), 3*time.Second)
+		t.Cleanup(cancel)
+
+		// generate a new token from the acceptor
+		req := pbpeering.GenerateTokenRequest{
+			PeerName: "my-peer-dialer",
+		}
+		resp, err := acceptorClient.GenerateToken(ctx, &req)
+		require.NoError(t, err)
+
+		token, err := acceptor.peeringBackend.DecodeToken([]byte(resp.PeeringToken))
+		require.NoError(t, err)
+
+		// we will update the token with bad addresses to assert it doesn't clobber existing ones
+		token.ServerAddresses = []string{"1.2.3.4:1234"}
+
+		badToken, err := acceptor.peeringBackend.EncodeToken(token)
+		require.NoError(t, err)
+
+		ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+		t.Cleanup(cancel)
+
+		// Try establishing.
+		// This call will only succeed if the bad address was not used in the calls to exchange the peering secret.
+		establishReq := pbpeering.EstablishRequest{
+			PeerName:     "my-peer-acceptor",
+			PeeringToken: string(badToken),
+		}
+		_, err = dialerClient.Establish(ctx, &establishReq)
+		require.NoError(t, err)
+
+		p, err := dialerClient.PeeringRead(ctx, &pbpeering.PeeringReadRequest{Name: "my-peer-acceptor"})
+		require.NoError(t, err)
+		require.NotContains(t, p.Peering.PeerServerAddresses, "1.2.3.4:1234")
+	})
+
+	testutil.RunStep(t, "updated server addresses are picked up by the leader", func(t *testing.T) {
+		// force close the acceptor's gRPC server so the dialier retries with a new address.
+		acceptor.externalGRPCServer.Stop()
+
+		clone := proto.Clone(p.Peering)
+		updated := clone.(*pbpeering.Peering)
+		// start with a bad address so we can assert for a specific error
+		updated.PeerServerAddresses = append([]string{
+			"bad",
+		}, p.Peering.PeerServerAddresses...)
+
+		// this write will wake up the watch on the leader to refetch server addresses
+		require.NoError(t, dialer.fsm.State().PeeringWrite(2000, &pbpeering.PeeringWriteRequest{Peering: updated}))
+
+		retry.Run(t, func(r *retry.R) {
+			status, found := dialer.peerStreamServer.StreamStatus(p.Peering.ID)
+			require.True(r, found)
+			// We assert for this error to be set which would indicate that we iterated
+			// through a bad address.
+			require.Contains(r, status.LastSendErrorMessage, "transport: Error while dialing dial tcp: address bad: missing port in address")
+			require.False(r, status.Connected)
+		})
+	})
+}

agent/consul/server.go

@@ -370,9 +370,9 @@ type Server struct {
 
 	// peerStreamServer is a server used to handle peering streams from external clusters.
 	peerStreamServer *peerstream.Server
+
 	// peeringServer handles peering RPC requests internal to this cluster, like generating peering tokens.
-	peeringServer     *peering.Server
+	peeringServer *peering.Server
-	peerStreamTracker *peerstream.Tracker
 
 	// embedded struct to hold all the enterprise specific data
 	EnterpriseServer
@@ -724,11 +724,9 @@ func NewServer(config *Config, flat Deps, externalGRPCServer *grpc.Server) (*Ser
 		Logger:      logger.Named("grpc-api.server-discovery"),
 	}).Register(s.externalGRPCServer)
 
-	s.peerStreamTracker = peerstream.NewTracker()
 	s.peeringBackend = NewPeeringBackend(s)
 	s.peerStreamServer = peerstream.NewServer(peerstream.Config{
 		Backend:        s.peeringBackend,
-		Tracker:        s.peerStreamTracker,
 		GetStore:       func() peerstream.StateStore { return s.FSM().State() },
 		Logger:         logger.Named("grpc-api.peerstream"),
 		ACLResolver:    s.ACLResolver,
@@ -790,7 +788,7 @@ func newGRPCHandlerFromConfig(deps Deps, config *Config, s *Server) connHandler
 
 	p := peering.NewServer(peering.Config{
 		Backend: s.peeringBackend,
-		Tracker: s.peerStreamTracker,
+		Tracker: s.peerStreamServer.Tracker,
 		Logger:  deps.Logger.Named("grpc-api.peering"),
 		ForwardRPC: func(info structs.RPCInfo, fn func(*grpc.ClientConn) error) (bool, error) {
 			// Only forward the request if the dc in the request matches the server's datacenter.
@@ -1574,12 +1572,12 @@ func (s *Server) Stats() map[string]map[string]string {
 // GetLANCoordinate returns the coordinate of the node in the LAN gossip
 // pool.
 //
-// - Clients return a single coordinate for the single gossip pool they are
+//   - Clients return a single coordinate for the single gossip pool they are
-//   in (default, segment, or partition).
+//     in (default, segment, or partition).
 //
-// - Servers return one coordinate for their canonical gossip pool (i.e.
+//   - Servers return one coordinate for their canonical gossip pool (i.e.
-//   default partition/segment) and one per segment they are also ancillary
+//     default partition/segment) and one per segment they are also ancillary
-//   members of.
+//     members of.
 //
 // NOTE: servers do not emit coordinates for partitioned gossip pools they
 // are ancillary members of.

agent/consul/state/catalog.go

@@ -1134,7 +1134,7 @@ func terminatingGatewayVirtualIPsSupported(tx ReadTxn, ws memdb.WatchSet) (bool,
 }
 
 // Services returns all services along with a list of associated tags.
-func (s *Store) Services(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta, peerName string) (uint64, structs.Services, error) {
+func (s *Store) Services(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta, peerName string) (uint64, []*structs.ServiceNode, error) {
 	tx := s.db.Txn(false)
 	defer tx.Abort()
 
@@ -1148,30 +1148,11 @@ func (s *Store) Services(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta, peerNam
 	}
 	ws.Add(services.WatchCh())
 
-	// Rip through the services and enumerate them and their unique set of
+	var result []*structs.ServiceNode
-	// tags.
-	unique := make(map[string]map[string]struct{})
 	for service := services.Next(); service != nil; service = services.Next() {
-		svc := service.(*structs.ServiceNode)
+		result = append(result, service.(*structs.ServiceNode))
-		tags, ok := unique[svc.ServiceName]
-		if !ok {
-			unique[svc.ServiceName] = make(map[string]struct{})
-			tags = unique[svc.ServiceName]
-		}
-		for _, tag := range svc.ServiceTags {
-			tags[tag] = struct{}{}
-		}
 	}
-
+	return idx, result, nil
-	// Generate the output structure.
-	var results = make(structs.Services)
-	for service, tags := range unique {
-		results[service] = make([]string, 0, len(tags))
-		for tag := range tags {
-			results[service] = append(results[service], tag)
-		}
-	}
-	return idx, results, nil
 }
 
 func (s *Store) ServiceList(ws memdb.WatchSet, entMeta *acl.EnterpriseMeta, peerName string) (uint64, structs.ServiceList, error) {
@@ -1212,7 +1193,7 @@ func serviceListTxn(tx ReadTxn, ws memdb.WatchSet, entMeta *acl.EnterpriseMeta,
 }
 
 // ServicesByNodeMeta returns all services, filtered by the given node metadata.
-func (s *Store) ServicesByNodeMeta(ws memdb.WatchSet, filters map[string]string, entMeta *acl.EnterpriseMeta, peerName string) (uint64, structs.Services, error) {
+func (s *Store) ServicesByNodeMeta(ws memdb.WatchSet, filters map[string]string, entMeta *acl.EnterpriseMeta, peerName string) (uint64, []*structs.ServiceNode, error) {
 	tx := s.db.Txn(false)
 	defer tx.Abort()
 
@@ -1259,8 +1240,7 @@ func (s *Store) ServicesByNodeMeta(ws memdb.WatchSet, filters map[string]string,
 	}
 	allServicesCh := allServices.WatchCh()
 
-	// Populate the services map
+	var result structs.ServiceNodes
-	unique := make(map[string]map[string]struct{})
 	for node := nodes.Next(); node != nil; node = nodes.Next() {
 		n := node.(*structs.Node)
 		if len(filters) > 1 && !structs.SatisfiesMetaFilters(n.Meta, filters) {
@@ -1274,30 +1254,11 @@ func (s *Store) ServicesByNodeMeta(ws memdb.WatchSet, filters map[string]string,
 		}
 		ws.AddWithLimit(watchLimit, services.WatchCh(), allServicesCh)
 
-		// Rip through the services and enumerate them and their unique set of
-		// tags.
 		for service := services.Next(); service != nil; service = services.Next() {
-			svc := service.(*structs.ServiceNode)
+			result = append(result, service.(*structs.ServiceNode))
-			tags, ok := unique[svc.ServiceName]
-			if !ok {
-				unique[svc.ServiceName] = make(map[string]struct{})
-				tags = unique[svc.ServiceName]
-			}
-			for _, tag := range svc.ServiceTags {
-				tags[tag] = struct{}{}
-			}
 		}
 	}
-
+	return idx, result, nil
-	// Generate the output structure.
-	var results = make(structs.Services)
-	for service, tags := range unique {
-		results[service] = make([]string, 0, len(tags))
-		for tag := range tags {
-			results[service] = append(results[service], tag)
-		}
-	}
-	return idx, results, nil
 }
 
 // maxIndexForService return the maximum Raft Index for a service
@@ -1717,6 +1678,9 @@ func (s *Store) ServiceNode(nodeID, nodeName, serviceID string, entMeta *acl.Ent
 	if err != nil {
 		return 0, nil, fmt.Errorf("failed querying service for node %q: %w", node.Node, err)
 	}
+	if service != nil {
+		service.ID = node.ID
+	}
 
 	return idx, service, nil
 }

agent/consul/state/catalog_test.go

@@ -12,6 +12,8 @@ import (
 
 	"github.com/hashicorp/consul/acl"
 
+	"github.com/google/go-cmp/cmp"
+	"github.com/google/go-cmp/cmp/cmpopts"
 	"github.com/hashicorp/go-memdb"
 	"github.com/hashicorp/go-uuid"
 	"github.com/stretchr/testify/assert"
@@ -270,17 +272,20 @@ func TestStateStore_EnsureRegistration(t *testing.T) {
 			require.Equal(t, uint64(2), idx)
 			require.Equal(t, svcmap["redis1"], r)
 
+			exp := svcmap["redis1"].ToServiceNode("node1")
+			exp.ID = nodeID
+
 			// lookup service by node name
 			idx, sn, err := s.ServiceNode("", "node1", "redis1", nil, peerName)
 			require.NoError(t, err)
 			require.Equal(t, uint64(2), idx)
-			require.Equal(t, svcmap["redis1"].ToServiceNode("node1"), sn)
+			require.Equal(t, exp, sn)
 
 			// lookup service by node ID
 			idx, sn, err = s.ServiceNode(string(nodeID), "", "redis1", nil, peerName)
 			require.NoError(t, err)
 			require.Equal(t, uint64(2), idx)
-			require.Equal(t, svcmap["redis1"].ToServiceNode("node1"), sn)
+			require.Equal(t, exp, sn)
 
 			// lookup service by invalid node
 			_, _, err = s.ServiceNode("", "invalid-node", "redis1", nil, peerName)
@@ -2102,10 +2107,13 @@ func TestStateStore_Services(t *testing.T) {
 		Address: "1.1.1.1",
 		Port:    1111,
 	}
+	ns1.EnterpriseMeta.Normalize()
 	if err := s.EnsureService(2, "node1", ns1); err != nil {
 		t.Fatalf("err: %s", err)
 	}
-	testRegisterService(t, s, 3, "node1", "dogs")
+	ns1Dogs := testRegisterService(t, s, 3, "node1", "dogs")
+	ns1Dogs.EnterpriseMeta.Normalize()
+
 	testRegisterNode(t, s, 4, "node2")
 	ns2 := &structs.NodeService{
 		ID:      "service3",
@@ -2114,6 +2122,7 @@ func TestStateStore_Services(t *testing.T) {
 		Address: "1.1.1.1",
 		Port:    1111,
 	}
+	ns2.EnterpriseMeta.Normalize()
 	if err := s.EnsureService(5, "node2", ns2); err != nil {
 		t.Fatalf("err: %s", err)
 	}
@@ -2131,19 +2140,13 @@ func TestStateStore_Services(t *testing.T) {
 		t.Fatalf("bad index: %d", idx)
 	}
 
-	// Verify the result. We sort the lists since the order is
+	// Verify the result.
-	// non-deterministic (it's built using a map internally).
+	expected := []*structs.ServiceNode{
-	expected := structs.Services{
+		ns1Dogs.ToServiceNode("node1"),
-		"redis": []string{"prod", "primary", "replica"},
+		ns1.ToServiceNode("node1"),
-		"dogs":  []string{},
+		ns2.ToServiceNode("node2"),
-	}
-	sort.Strings(expected["redis"])
-	for _, tags := range services {
-		sort.Strings(tags)
-	}
-	if !reflect.DeepEqual(expected, services) {
-		t.Fatalf("bad: %#v", services)
 	}
+	assertDeepEqual(t, expected, services, cmpopts.IgnoreFields(structs.ServiceNode{}, "RaftIndex"))
 
 	// Deleting a node with a service should fire the watch.
 	if err := s.DeleteNode(6, "node1", nil, ""); err != nil {
@@ -2182,6 +2185,7 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		Address: "1.1.1.1",
 		Port:    1111,
 	}
+	ns1.EnterpriseMeta.Normalize()
 	if err := s.EnsureService(2, "node0", ns1); err != nil {
 		t.Fatalf("err: %s", err)
 	}
@@ -2192,6 +2196,7 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		Address: "1.1.1.1",
 		Port:    1111,
 	}
+	ns2.EnterpriseMeta.Normalize()
 	if err := s.EnsureService(3, "node1", ns2); err != nil {
 		t.Fatalf("err: %s", err)
 	}
@@ -2206,11 +2211,10 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		if err != nil {
 			t.Fatalf("err: %s", err)
 		}
-		expected := structs.Services{
+		expected := []*structs.ServiceNode{
-			"redis": []string{"primary", "prod"},
+			ns1.ToServiceNode("node0"),
 		}
-		sort.Strings(res["redis"])
+		assertDeepEqual(t, res, expected, cmpopts.IgnoreFields(structs.ServiceNode{}, "RaftIndex"))
-		require.Equal(t, expected, res)
 	})
 
 	t.Run("Get all services using the common meta value", func(t *testing.T) {
@@ -2218,11 +2222,12 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		if err != nil {
 			t.Fatalf("err: %s", err)
 		}
-		expected := structs.Services{
+		require.Len(t, res, 2)
-			"redis": []string{"primary", "prod", "replica"},
+		expected := []*structs.ServiceNode{
+			ns1.ToServiceNode("node0"),
+			ns2.ToServiceNode("node1"),
 		}
-		sort.Strings(res["redis"])
+		assertDeepEqual(t, res, expected, cmpopts.IgnoreFields(structs.ServiceNode{}, "RaftIndex"))
-		require.Equal(t, expected, res)
 	})
 
 	t.Run("Get an empty list for an invalid meta value", func(t *testing.T) {
@@ -2230,8 +2235,8 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		if err != nil {
 			t.Fatalf("err: %s", err)
 		}
-		expected := structs.Services{}
+		var expected []*structs.ServiceNode
-		require.Equal(t, expected, res)
+		assertDeepEqual(t, res, expected, cmpopts.IgnoreFields(structs.ServiceNode{}, "RaftIndex"))
 	})
 
 	t.Run("Get the first node's service instance using multiple meta filters", func(t *testing.T) {
@@ -2239,11 +2244,10 @@ func TestStateStore_ServicesByNodeMeta(t *testing.T) {
 		if err != nil {
 			t.Fatalf("err: %s", err)
 		}
-		expected := structs.Services{
+		expected := []*structs.ServiceNode{
-			"redis": []string{"primary", "prod"},
+			ns1.ToServiceNode("node0"),
 		}
-		sort.Strings(res["redis"])
+		assertDeepEqual(t, res, expected, cmpopts.IgnoreFields(structs.ServiceNode{}, "RaftIndex"))
-		require.Equal(t, expected, res)
 	})
 
 	t.Run("Registering some unrelated node + service should not fire the watch.", func(t *testing.T) {
@@ -8807,3 +8811,10 @@ func setVirtualIPFlags(t *testing.T, s *Store) {
 		Value: "true",
 	}))
 }
+
+func assertDeepEqual(t *testing.T, x, y interface{}, opts ...cmp.Option) {
+	t.Helper()
+	if diff := cmp.Diff(x, y, opts...); diff != "" {
+		t.Fatalf("assertion failed: values are not equal\n--- expected\n+++ actual\n%v", diff)
+	}
+}

agent/consul/state/peering.go

@@ -7,12 +7,13 @@ import (
 	"strings"
 
 	"github.com/golang/protobuf/proto"
+	"github.com/hashicorp/go-memdb"
+
 	"github.com/hashicorp/consul/acl"
 	"github.com/hashicorp/consul/agent/configentry"
 	"github.com/hashicorp/consul/agent/structs"
 	"github.com/hashicorp/consul/lib/maps"
 	"github.com/hashicorp/consul/proto/pbpeering"
-	"github.com/hashicorp/go-memdb"
 )
 
 const (
@@ -534,6 +535,12 @@ func (s *Store) PeeringWrite(idx uint64, req *pbpeering.PeeringWriteRequest) err
 	if req.Peering.Name == "" {
 		return errors.New("Missing Peering Name")
 	}
+	if req.Peering.State == pbpeering.PeeringState_DELETING && (req.Peering.DeletedAt == nil || structs.IsZeroProtoTime(req.Peering.DeletedAt)) {
+		return errors.New("Missing deletion time for peering in deleting state")
+	}
+	if req.Peering.DeletedAt != nil && !structs.IsZeroProtoTime(req.Peering.DeletedAt) && req.Peering.State != pbpeering.PeeringState_DELETING {
+		return fmt.Errorf("Unexpected state for peering with deletion time: %s", pbpeering.PeeringStateToAPI(req.Peering.State))
+	}
 
 	// Ensure the name is unique (cannot conflict with another peering with a different ID).
 	_, existing, err := peeringReadTxn(tx, nil, Query{
@@ -545,11 +552,32 @@ func (s *Store) PeeringWrite(idx uint64, req *pbpeering.PeeringWriteRequest) err
 	}
 
 	if existing != nil {
+		if req.Peering.ShouldDial() != existing.ShouldDial() {
+			return fmt.Errorf("Cannot switch peering dialing mode from %t to %t", existing.ShouldDial(), req.Peering.ShouldDial())
+		}
+
 		if req.Peering.ID != existing.ID {
 			return fmt.Errorf("A peering already exists with the name %q and a different ID %q", req.Peering.Name, existing.ID)
 		}
+
+		// Nothing to do if our peer wants to terminate the peering but the peering is already marked for deletion.
+		if existing.State == pbpeering.PeeringState_DELETING && req.Peering.State == pbpeering.PeeringState_TERMINATED {
+			return nil
+		}
+
+		// No-op deletion
+		if existing.State == pbpeering.PeeringState_DELETING && req.Peering.State == pbpeering.PeeringState_DELETING {
+			return nil
+		}
+
+		// No-op termination
+		if existing.State == pbpeering.PeeringState_TERMINATED && req.Peering.State == pbpeering.PeeringState_TERMINATED {
+			return nil
+		}
+
 		// Prevent modifications to Peering marked for deletion.
-		if !existing.IsActive() {
+		// This blocks generating new peering tokens or re-establishing the peering until the peering is done deleting.
+		if existing.State == pbpeering.PeeringState_DELETING {
 			return fmt.Errorf("cannot write to peering that is marked for deletion")
 		}
 
@@ -581,8 +609,8 @@ func (s *Store) PeeringWrite(idx uint64, req *pbpeering.PeeringWriteRequest) err
 		req.Peering.ModifyIndex = idx
 	}
 
-	// Ensure associated secrets are cleaned up when a peering is marked for deletion.
+	// Ensure associated secrets are cleaned up when a peering is marked for deletion or terminated.
-	if req.Peering.State == pbpeering.PeeringState_DELETING {
+	if !req.Peering.IsActive() {
 		if err := peeringSecretsDeleteTxn(tx, req.Peering.ID, req.Peering.ShouldDial()); err != nil {
 			return fmt.Errorf("failed to delete peering secrets: %w", err)
 		}
@@ -981,7 +1009,7 @@ func peeringsForServiceTxn(tx ReadTxn, ws memdb.WatchSet, serviceName string, en
 		if idx > maxIdx {
 			maxIdx = idx
 		}
-		if peering == nil || !peering.IsActive() {
+		if !peering.IsActive() {
 			continue
 		}
 		peerings = append(peerings, peering)

agent/consul/state/peering_test.go

@@ -950,6 +950,7 @@ func TestStore_Peering_Watch(t *testing.T) {
 			Peering: &pbpeering.Peering{
 				ID:        testFooPeerID,
 				Name:      "foo",
+				State:     pbpeering.PeeringState_DELETING,
 				DeletedAt: structs.TimeToProto(time.Now()),
 			},
 		})
@@ -976,6 +977,7 @@ func TestStore_Peering_Watch(t *testing.T) {
 		err := s.PeeringWrite(lastIdx, &pbpeering.PeeringWriteRequest{Peering: &pbpeering.Peering{
 			ID:        testBarPeerID,
 			Name:      "bar",
+			State:     pbpeering.PeeringState_DELETING,
 			DeletedAt: structs.TimeToProto(time.Now()),
 		},
 		})
@@ -1077,6 +1079,7 @@ func TestStore_PeeringList_Watch(t *testing.T) {
 			Peering: &pbpeering.Peering{
 				ID:        testFooPeerID,
 				Name:      "foo",
+				State:     pbpeering.PeeringState_DELETING,
 				DeletedAt: structs.TimeToProto(time.Now()),
 				Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 			},
@@ -1112,16 +1115,22 @@ func TestStore_PeeringWrite(t *testing.T) {
 	// Each case depends on the previous.
 	s := NewStateStore(nil)
 
+	testTime := time.Now()
+
+	type expectations struct {
+		peering *pbpeering.Peering
+		secrets *pbpeering.PeeringSecrets
+		err     string
+	}
 	type testcase struct {
-		name          string
+		name   string
-		input         *pbpeering.PeeringWriteRequest
+		input  *pbpeering.PeeringWriteRequest
-		expectSecrets *pbpeering.PeeringSecrets
+		expect expectations
-		expectErr     string
 	}
 	run := func(t *testing.T, tc testcase) {
 		err := s.PeeringWrite(10, tc.input)
-		if tc.expectErr != "" {
+		if tc.expect.err != "" {
-			testutil.RequireErrorContains(t, err, tc.expectErr)
+			testutil.RequireErrorContains(t, err, tc.expect.err)
 			return
 		}
 		require.NoError(t, err)
@@ -1133,52 +1142,176 @@ func TestStore_PeeringWrite(t *testing.T) {
 		_, p, err := s.PeeringRead(nil, q)
 		require.NoError(t, err)
 		require.NotNil(t, p)
-		require.Equal(t, tc.input.Peering.State, p.State)
+		require.Equal(t, tc.expect.peering.State, p.State)
-		require.Equal(t, tc.input.Peering.Name, p.Name)
+		require.Equal(t, tc.expect.peering.Name, p.Name)
+		require.Equal(t, tc.expect.peering.Meta, p.Meta)
+		if tc.expect.peering.DeletedAt != nil {
+			require.Equal(t, tc.expect.peering.DeletedAt, p.DeletedAt)
+		}
 
 		secrets, err := s.PeeringSecretsRead(nil, tc.input.Peering.ID)
 		require.NoError(t, err)
-		prototest.AssertDeepEqual(t, tc.expectSecrets, secrets)
+		prototest.AssertDeepEqual(t, tc.expect.secrets, secrets)
 	}
 	tcs := []testcase{
 		{
 			name: "create baz",
 			input: &pbpeering.PeeringWriteRequest{
 				Peering: &pbpeering.Peering{
-					ID:        testBazPeerID,
+					ID:                  testBazPeerID,
-					Name:      "baz",
+					Name:                "baz",
-					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+					State:               pbpeering.PeeringState_ESTABLISHING,
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 				},
 				SecretsRequest: &pbpeering.SecretsWriteRequest{
 					PeerID: testBazPeerID,
-					Request: &pbpeering.SecretsWriteRequest_GenerateToken{
+					Request: &pbpeering.SecretsWriteRequest_Establish{
-						GenerateToken: &pbpeering.SecretsWriteRequest_GenerateTokenRequest{
+						Establish: &pbpeering.SecretsWriteRequest_EstablishRequest{
-							EstablishmentSecret: testBazSecretID,
+							ActiveStreamSecret: testBazSecretID,
 						},
 					},
 				},
 			},
-			expectSecrets: &pbpeering.PeeringSecrets{
+			expect: expectations{
-				PeerID: testBazPeerID,
+				peering: &pbpeering.Peering{
-				Establishment: &pbpeering.PeeringSecrets_Establishment{
+					ID:    testBazPeerID,
-					SecretID: testBazSecretID,
+					Name:  "baz",
+					State: pbpeering.PeeringState_ESTABLISHING,
 				},
+				secrets: &pbpeering.PeeringSecrets{
+					PeerID: testBazPeerID,
+					Stream: &pbpeering.PeeringSecrets_Stream{
+						ActiveSecretID: testBazSecretID,
+					},
+				},
+			},
+		},
+		{
+			name: "cannot change ID for baz",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:                  "123",
+					Name:                "baz",
+					State:               pbpeering.PeeringState_FAILING,
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				err: `A peering already exists with the name "baz" and a different ID`,
+			},
+		},
+		{
+			name: "cannot change dialer status for baz",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:    "123",
+					Name:  "baz",
+					State: pbpeering.PeeringState_FAILING,
+					// Excluding the peer server addresses leads to baz not being considered a dialer.
+					// PeerServerAddresses: []string{"localhost:8502"},
+					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				err: "Cannot switch peering dialing mode from true to false",
 			},
 		},
 		{
 			name: "update baz",
 			input: &pbpeering.PeeringWriteRequest{
 				Peering: &pbpeering.Peering{
-					ID:        testBazPeerID,
+					ID:                  testBazPeerID,
-					Name:      "baz",
+					Name:                "baz",
-					State:     pbpeering.PeeringState_FAILING,
+					State:               pbpeering.PeeringState_FAILING,
-					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 				},
 			},
-			expectSecrets: &pbpeering.PeeringSecrets{
+			expect: expectations{
-				PeerID: testBazPeerID,
+				peering: &pbpeering.Peering{
-				Establishment: &pbpeering.PeeringSecrets_Establishment{
+					ID:    testBazPeerID,
-					SecretID: testBazSecretID,
+					Name:  "baz",
+					State: pbpeering.PeeringState_FAILING,
+				},
+				secrets: &pbpeering.PeeringSecrets{
+					PeerID: testBazPeerID,
+					Stream: &pbpeering.PeeringSecrets_Stream{
+						ActiveSecretID: testBazSecretID,
+					},
+				},
+			},
+		},
+		{
+			name: "if no state was included in request it is inherited from existing",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:   testBazPeerID,
+					Name: "baz",
+					// Send undefined state.
+					// State:               pbpeering.PeeringState_FAILING,
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				peering: &pbpeering.Peering{
+					ID:   testBazPeerID,
+					Name: "baz",
+					// Previous failing state is picked up.
+					State: pbpeering.PeeringState_FAILING,
+				},
+				secrets: &pbpeering.PeeringSecrets{
+					PeerID: testBazPeerID,
+					Stream: &pbpeering.PeeringSecrets_Stream{
+						ActiveSecretID: testBazSecretID,
+					},
+				},
+			},
+		},
+		{
+			name: "mark baz as terminated",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:                  testBazPeerID,
+					Name:                "baz",
+					State:               pbpeering.PeeringState_TERMINATED,
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				peering: &pbpeering.Peering{
+					ID:    testBazPeerID,
+					Name:  "baz",
+					State: pbpeering.PeeringState_TERMINATED,
+				},
+				// Secrets for baz should have been deleted
+				secrets: nil,
+			},
+		},
+		{
+			name: "cannot modify peering during no-op termination",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:                  testBazPeerID,
+					Name:                "baz",
+					State:               pbpeering.PeeringState_TERMINATED,
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+					PeerServerAddresses: []string{"localhost:8502"},
+
+					// Attempt to add metadata
+					Meta: map[string]string{"foo": "bar"},
+				},
+			},
+			expect: expectations{
+				peering: &pbpeering.Peering{
+					ID:    testBazPeerID,
+					Name:  "baz",
+					State: pbpeering.PeeringState_TERMINATED,
+					// Meta should be unchanged.
+					Meta: nil,
 				},
 			},
 		},
@@ -1186,42 +1319,104 @@ func TestStore_PeeringWrite(t *testing.T) {
 			name: "mark baz for deletion",
 			input: &pbpeering.PeeringWriteRequest{
 				Peering: &pbpeering.Peering{
+					ID:                  testBazPeerID,
+					Name:                "baz",
+					State:               pbpeering.PeeringState_DELETING,
+					PeerServerAddresses: []string{"localhost:8502"},
+					DeletedAt:           structs.TimeToProto(testTime),
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				peering: &pbpeering.Peering{
 					ID:        testBazPeerID,
 					Name:      "baz",
 					State:     pbpeering.PeeringState_DELETING,
-					DeletedAt: structs.TimeToProto(time.Now()),
+					DeletedAt: structs.TimeToProto(testTime),
-					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+				secrets: nil,
+			},
+		},
+		{
+			name: "deleting a deleted peering is a no-op",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:                  testBazPeerID,
+					Name:                "baz",
+					State:               pbpeering.PeeringState_DELETING,
+					PeerServerAddresses: []string{"localhost:8502"},
+					DeletedAt:           structs.TimeToProto(time.Now()),
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 				},
 			},
-			// Secrets for baz should have been deleted
+			expect: expectations{
-			expectSecrets: nil,
+				peering: &pbpeering.Peering{
+					ID:   testBazPeerID,
+					Name: "baz",
+					// Still marked as deleting at the original testTime
+					State:     pbpeering.PeeringState_DELETING,
+					DeletedAt: structs.TimeToProto(testTime),
+				},
+				// Secrets for baz should have been deleted
+				secrets: nil,
+			},
+		},
+		{
+			name: "terminating a peering marked for deletion is a no-op",
+			input: &pbpeering.PeeringWriteRequest{
+				Peering: &pbpeering.Peering{
+					ID:                  testBazPeerID,
+					Name:                "baz",
+					State:               pbpeering.PeeringState_TERMINATED,
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+				},
+			},
+			expect: expectations{
+				peering: &pbpeering.Peering{
+					ID:   testBazPeerID,
+					Name: "baz",
+					// Still marked as deleting
+					State: pbpeering.PeeringState_DELETING,
+				},
+				// Secrets for baz should have been deleted
+				secrets: nil,
+			},
 		},
 		{
 			name: "cannot update peering marked for deletion",
 			input: &pbpeering.PeeringWriteRequest{
 				Peering: &pbpeering.Peering{
-					ID:   testBazPeerID,
+					ID:                  testBazPeerID,
-					Name: "baz",
+					Name:                "baz",
+					PeerServerAddresses: []string{"localhost:8502"},
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+
 					// Attempt to add metadata
 					Meta: map[string]string{
 						"source": "kubernetes",
 					},
-					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 				},
 			},
-			expectErr: "cannot write to peering that is marked for deletion",
+			expect: expectations{
+				err: "cannot write to peering that is marked for deletion",
+			},
 		},
 		{
 			name: "cannot create peering marked for deletion",
 			input: &pbpeering.PeeringWriteRequest{
 				Peering: &pbpeering.Peering{
-					ID:        testFooPeerID,
+					ID:                  testFooPeerID,
-					Name:      "foo",
+					Name:                "foo",
-					DeletedAt: structs.TimeToProto(time.Now()),
+					PeerServerAddresses: []string{"localhost:8502"},
-					Partition: structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
+					State:               pbpeering.PeeringState_DELETING,
+					DeletedAt:           structs.TimeToProto(time.Now()),
+					Partition:           structs.NodeEnterpriseMetaInDefaultPartition().PartitionOrEmpty(),
 				},
 			},
-			expectErr: "cannot create a new peering marked for deletion",
+			expect: expectations{
+				err: "cannot create a new peering marked for deletion",
+			},
 		},
 	}
 	for _, tc := range tcs {
@@ -1246,6 +1441,7 @@ func TestStore_PeeringDelete(t *testing.T) {
 			Peering: &pbpeering.Peering{
 				ID:        testFooPeerID,
 				Name:      "foo",
+				State:     pbpeering.PeeringState_DELETING,
 				DeletedAt: structs.TimeToProto(time.Now()),
 			},
 		}))
@@ -1461,7 +1657,13 @@ func TestStateStore_ExportedServicesForPeer(t *testing.T) {
 	}
 
 	newTarget := func(service, serviceSubset, datacenter string) *structs.DiscoveryTarget {
-		t := structs.NewDiscoveryTarget(service, serviceSubset, "default", "default", datacenter)
+		t := structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:       service,
+			ServiceSubset: serviceSubset,
+			Partition:     "default",
+			Namespace:     "default",
+			Datacenter:    datacenter,
+		})
 		t.SNI = connect.TargetSNI(t, connect.TestTrustDomain)
 		t.Name = t.SNI
 		t.ConnectTimeout = 5 * time.Second // default
@@ -1753,6 +1955,7 @@ func TestStateStore_PeeringsForService(t *testing.T) {
 				copied := pbpeering.Peering{
 					ID:        tp.peering.ID,
 					Name:      tp.peering.Name,
+					State:     pbpeering.PeeringState_DELETING,
 					DeletedAt: structs.TimeToProto(time.Now()),
 				}
 				require.NoError(t, s.PeeringWrite(lastIdx, &pbpeering.PeeringWriteRequest{Peering: &copied}))
@@ -2195,6 +2398,7 @@ func TestStore_TrustBundleListByService(t *testing.T) {
 			Peering: &pbpeering.Peering{
 				ID:        peerID1,
 				Name:      "peer1",
+				State:     pbpeering.PeeringState_DELETING,
 				DeletedAt: structs.TimeToProto(time.Now()),
 			},
 		}))

agent/consul/state/state_store_test.go

@@ -146,13 +146,13 @@ func testRegisterServiceOpts(t *testing.T, s *Store, idx uint64, nodeID, service
 // testRegisterServiceWithChange registers a service and allow ensuring the consul index is updated
 // even if service already exists if using `modifyAccordingIndex`.
 // This is done by setting the transaction ID in "version" meta so service will be updated if it already exists
-func testRegisterServiceWithChange(t *testing.T, s *Store, idx uint64, nodeID, serviceID string, modifyAccordingIndex bool) {
+func testRegisterServiceWithChange(t *testing.T, s *Store, idx uint64, nodeID, serviceID string, modifyAccordingIndex bool) *structs.NodeService {
-	testRegisterServiceWithChangeOpts(t, s, idx, nodeID, serviceID, modifyAccordingIndex)
+	return testRegisterServiceWithChangeOpts(t, s, idx, nodeID, serviceID, modifyAccordingIndex)
 }
 
 // testRegisterServiceWithChangeOpts is the same as testRegisterServiceWithChange with the addition of opts that can
 // modify the service prior to writing.
-func testRegisterServiceWithChangeOpts(t *testing.T, s *Store, idx uint64, nodeID, serviceID string, modifyAccordingIndex bool, opts ...func(service *structs.NodeService)) {
+func testRegisterServiceWithChangeOpts(t *testing.T, s *Store, idx uint64, nodeID, serviceID string, modifyAccordingIndex bool, opts ...func(service *structs.NodeService)) *structs.NodeService {
 	meta := make(map[string]string)
 	if modifyAccordingIndex {
 		meta["version"] = fmt.Sprint(idx)
@@ -183,14 +183,15 @@ func testRegisterServiceWithChangeOpts(t *testing.T, s *Store, idx uint64, nodeI
 		result.ServiceID != serviceID {
 		t.Fatalf("bad service: %#v", result)
 	}
+	return svc
 }
 
 // testRegisterService register a service with given transaction idx
 // If the service already exists, transaction number might not be increased
 // Use `testRegisterServiceWithChange()` if you want perform a registration that
 // ensures the transaction is updated by setting idx in Meta of Service
-func testRegisterService(t *testing.T, s *Store, idx uint64, nodeID, serviceID string) {
+func testRegisterService(t *testing.T, s *Store, idx uint64, nodeID, serviceID string) *structs.NodeService {
-	testRegisterServiceWithChange(t, s, idx, nodeID, serviceID, false)
+	return testRegisterServiceWithChange(t, s, idx, nodeID, serviceID, false)
 }
 
 func testRegisterConnectService(t *testing.T, s *Store, idx uint64, nodeID, serviceID string) {

agent/consul/usagemetrics/usagemetrics.go

@@ -41,8 +41,8 @@ var Gauges = []prometheus.GaugeDefinition{
 		Help: "Measures the current number of server agents registered with Consul. It is only emitted by Consul servers. Added in v1.9.6.",
 	},
 	{
-		Name: []string{"consul", "kv", "entries"},
+		Name: []string{"consul", "state", "kv_entries"},
-		Help: "Measures the current number of server agents registered with Consul. It is only emitted by Consul servers. Added in v1.10.3.",
+		Help: "Measures the current number of entries in the Consul KV store. It is only emitted by Consul servers. Added in v1.10.3.",
 	},
 	{
 		Name: []string{"consul", "state", "connect_instances"},

agent/discovery_chain_endpoint_test.go

@@ -27,8 +27,17 @@ func TestDiscoveryChainRead(t *testing.T) {
 	defer a.Shutdown()
 	testrpc.WaitForTestAgent(t, a.RPC, "dc1")
 
-	newTarget := func(service, serviceSubset, namespace, partition, datacenter string) *structs.DiscoveryTarget {
+	newTarget := func(opts structs.DiscoveryTargetOpts) *structs.DiscoveryTarget {
-		t := structs.NewDiscoveryTarget(service, serviceSubset, namespace, partition, datacenter)
+		if opts.Namespace == "" {
+			opts.Namespace = "default"
+		}
+		if opts.Partition == "" {
+			opts.Partition = "default"
+		}
+		if opts.Datacenter == "" {
+			opts.Datacenter = "dc1"
+		}
+		t := structs.NewDiscoveryTarget(opts)
 		t.SNI = connect.TargetSNI(t, connect.TestClusterID+".consul")
 		t.Name = t.SNI
 		t.ConnectTimeout = 5 * time.Second // default
@@ -99,7 +108,7 @@ func TestDiscoveryChainRead(t *testing.T) {
 					},
 				},
 				Targets: map[string]*structs.DiscoveryTarget{
-					"web.default.default.dc1": newTarget("web", "", "default", "default", "dc1"),
+					"web.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 				},
 			}
 			require.Equal(t, expect, value.Chain)
@@ -144,7 +153,7 @@ func TestDiscoveryChainRead(t *testing.T) {
 					},
 				},
 				Targets: map[string]*structs.DiscoveryTarget{
-					"web.default.default.dc2": newTarget("web", "", "default", "default", "dc2"),
+					"web.default.default.dc2": newTarget(structs.DiscoveryTargetOpts{Service: "web", Datacenter: "dc2"}),
 				},
 			}
 			require.Equal(t, expect, value.Chain)
@@ -198,7 +207,7 @@ func TestDiscoveryChainRead(t *testing.T) {
 					},
 				},
 				Targets: map[string]*structs.DiscoveryTarget{
-					"web.default.default.dc1": newTarget("web", "", "default", "default", "dc1"),
+					"web.default.default.dc1": newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 				},
 			}
 			require.Equal(t, expect, value.Chain)
@@ -264,11 +273,11 @@ func TestDiscoveryChainRead(t *testing.T) {
 				},
 				Targets: map[string]*structs.DiscoveryTarget{
 					"web.default.default.dc1": targetWithConnectTimeout(
-						newTarget("web", "", "default", "default", "dc1"),
+						newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 						33*time.Second,
 					),
 					"web.default.default.dc2": targetWithConnectTimeout(
-						newTarget("web", "", "default", "default", "dc2"),
+						newTarget(structs.DiscoveryTargetOpts{Service: "web", Datacenter: "dc2"}),
 						33*time.Second,
 					),
 				},
@@ -280,7 +289,7 @@ func TestDiscoveryChainRead(t *testing.T) {
 	}))
 
 	expectTarget_DC1 := targetWithConnectTimeout(
-		newTarget("web", "", "default", "default", "dc1"),
+		newTarget(structs.DiscoveryTargetOpts{Service: "web"}),
 		22*time.Second,
 	)
 	expectTarget_DC1.MeshGateway = structs.MeshGatewayConfig{
@@ -288,7 +297,7 @@ func TestDiscoveryChainRead(t *testing.T) {
 	}
 
 	expectTarget_DC2 := targetWithConnectTimeout(
-		newTarget("web", "", "default", "default", "dc2"),
+		newTarget(structs.DiscoveryTargetOpts{Service: "web", Datacenter: "dc2"}),
 		22*time.Second,
 	)
 	expectTarget_DC2.MeshGateway = structs.MeshGatewayConfig{

agent/grpc-external/server.go

@@ -1,12 +1,13 @@
 package external
 
 import (
+	"time"
+
 	middleware "github.com/grpc-ecosystem/go-grpc-middleware"
 	recovery "github.com/grpc-ecosystem/go-grpc-middleware/recovery"
 	"google.golang.org/grpc"
 	"google.golang.org/grpc/credentials"
 	"google.golang.org/grpc/keepalive"
-	"time"
 
 	agentmiddleware "github.com/hashicorp/consul/agent/grpc-middleware"
 	"github.com/hashicorp/consul/tlsutil"
@@ -34,7 +35,7 @@ func NewServer(logger agentmiddleware.Logger, tls *tlsutil.Configurator) *grpc.S
 			MinTime: 15 * time.Second,
 		}),
 	}
-	if tls != nil && tls.GRPCTLSConfigured() {
+	if tls != nil && tls.GRPCServerUseTLS() {
 		creds := credentials.NewTLS(tls.IncomingGRPCConfig())
 		opts = append(opts, grpc.Creds(creds))
 	}

agent/grpc-external/services/dataplane/get_envoy_bootstrap_params.go

@@ -52,13 +52,21 @@ func (s *Server) GetEnvoyBootstrapParams(ctx context.Context, req *pbdataplane.G
 	}
 
 	// Build out the response
+	var serviceName string
+	if svc.ServiceKind == structs.ServiceKindConnectProxy {
+		serviceName = svc.ServiceProxy.DestinationServiceName
+	} else {
+		serviceName = svc.ServiceName
+	}
 
 	resp := &pbdataplane.GetEnvoyBootstrapParamsResponse{
-		Service:     svc.ServiceProxy.DestinationServiceName,
+		Service:     serviceName,
 		Partition:   svc.EnterpriseMeta.PartitionOrDefault(),
 		Namespace:   svc.EnterpriseMeta.NamespaceOrDefault(),
 		Datacenter:  s.Datacenter,
 		ServiceKind: convertToResponseServiceKind(svc.ServiceKind),
+		NodeName:    svc.Node,
+		NodeId:      string(svc.ID),
 	}
 
 	bootstrapConfig, err := structpb.NewStruct(svc.ServiceProxy.Config)

agent/grpc-external/services/dataplane/get_envoy_bootstrap_params_test.go

@@ -97,14 +97,20 @@ func TestGetEnvoyBootstrapParams_Success(t *testing.T) {
 		resp, err := client.GetEnvoyBootstrapParams(ctx, req)
 		require.NoError(t, err)
 
-		require.Equal(t, tc.registerReq.Service.Proxy.DestinationServiceName, resp.Service)
+		if tc.registerReq.Service.IsGateway() {
+			require.Equal(t, tc.registerReq.Service.Service, resp.Service)
+		} else {
+			require.Equal(t, tc.registerReq.Service.Proxy.DestinationServiceName, resp.Service)
+		}
+
 		require.Equal(t, serverDC, resp.Datacenter)
 		require.Equal(t, tc.registerReq.EnterpriseMeta.PartitionOrDefault(), resp.Partition)
 		require.Equal(t, tc.registerReq.EnterpriseMeta.NamespaceOrDefault(), resp.Namespace)
 		require.Contains(t, resp.Config.Fields, proxyConfigKey)
 		require.Equal(t, structpb.NewStringValue(proxyConfigValue), resp.Config.Fields[proxyConfigKey])
 		require.Equal(t, convertToResponseServiceKind(tc.registerReq.Service.Kind), resp.ServiceKind)
-
+		require.Equal(t, tc.registerReq.Node, resp.NodeName)
+		require.Equal(t, string(tc.registerReq.ID), resp.NodeId)
 	}
 
 	testCases := []testCase{

agent/grpc-external/services/peerstream/server.go

@@ -26,11 +26,12 @@ const (
 
 type Server struct {
 	Config
+
+	Tracker *Tracker
 }
 
 type Config struct {
 	Backend     Backend
-	Tracker     *Tracker
 	GetStore    func() StateStore
 	Logger      hclog.Logger
 	ForwardRPC  func(structs.RPCInfo, func(*grpc.ClientConn) error) (bool, error)
@@ -53,7 +54,6 @@ type ACLResolver interface {
 
 func NewServer(cfg Config) *Server {
 	requireNotNil(cfg.Backend, "Backend")
-	requireNotNil(cfg.Tracker, "Tracker")
 	requireNotNil(cfg.GetStore, "GetStore")
 	requireNotNil(cfg.Logger, "Logger")
 	// requireNotNil(cfg.ACLResolver, "ACLResolver") // TODO(peering): reenable check when ACLs are required
@@ -67,7 +67,8 @@ func NewServer(cfg Config) *Server {
 		cfg.incomingHeartbeatTimeout = defaultIncomingHeartbeatTimeout
 	}
 	return &Server{
-		Config: cfg,
+		Config:  cfg,
+		Tracker: NewTracker(cfg.incomingHeartbeatTimeout),
 	}
 }
 

agent/grpc-external/services/peerstream/stream_resources.go

@@ -575,6 +575,7 @@ func (s *Server) realHandleStream(streamReq HandleStreamRequest) error {
 					status.TrackRecvResourceSuccess()
 				}
 
+				// We are replying ACK or NACK depending on whether we successfully processed the response.
 				if err := streamSend(reply); err != nil {
 					return fmt.Errorf("failed to send to stream: %v", err)
 				}

agent/grpc-external/services/peerstream/stream_test.go

@@ -499,9 +499,8 @@ func TestStreamResources_Server_Terminate(t *testing.T) {
 		base: time.Date(2000, time.January, 1, 0, 0, 0, 0, time.UTC),
 	}
 
-	srv, store := newTestServer(t, func(c *Config) {
+	srv, store := newTestServer(t, nil)
-		c.Tracker.SetClock(it.Now)
+	srv.Tracker.setClock(it.Now)
-	})
 
 	p := writePeeringToBeDialed(t, store, 1, "my-peer")
 	require.Empty(t, p.PeerID, "should be empty if being dialed")
@@ -552,9 +551,8 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 		base: time.Date(2000, time.January, 1, 0, 0, 0, 0, time.UTC),
 	}
 
-	srv, store := newTestServer(t, func(c *Config) {
+	srv, store := newTestServer(t, nil)
-		c.Tracker.SetClock(it.Now)
+	srv.Tracker.setClock(it.Now)
-	})
 
 	// Set the initial roots and CA configuration.
 	_, rootA := writeInitialRootsAndCA(t, store)
@@ -572,7 +570,7 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 		})
 	})
 
-	var lastSendSuccess time.Time
+	var lastSendAck time.Time
 
 	testutil.RunStep(t, "ack tracked as success", func(t *testing.T) {
 		ack := &pbpeerstream.ReplicationMessage{
@@ -587,19 +585,19 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 			},
 		}
 
-		lastSendSuccess = it.FutureNow(1)
+		lastSendAck = it.FutureNow(1)
 		err := client.Send(ack)
 		require.NoError(t, err)
 
 		expect := Status{
 			Connected: true,
-			LastAck:   lastSendSuccess,
+			LastAck:   lastSendAck,
 		}
 
 		retry.Run(t, func(r *retry.R) {
-			status, ok := srv.StreamStatus(testPeerID)
+			rStatus, ok := srv.StreamStatus(testPeerID)
 			require.True(r, ok)
-			require.Equal(r, expect, status)
+			require.Equal(r, expect, rStatus)
 		})
 	})
 
@@ -629,15 +627,15 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 
 		expect := Status{
 			Connected:       true,
-			LastAck:         lastSendSuccess,
+			LastAck:         lastSendAck,
 			LastNack:        lastNack,
 			LastNackMessage: lastNackMsg,
 		}
 
 		retry.Run(t, func(r *retry.R) {
-			status, ok := srv.StreamStatus(testPeerID)
+			rStatus, ok := srv.StreamStatus(testPeerID)
 			require.True(r, ok)
-			require.Equal(r, expect, status)
+			require.Equal(r, expect, rStatus)
 		})
 	})
 
@@ -694,7 +692,7 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 
 		expect := Status{
 			Connected:               true,
-			LastAck:                 lastSendSuccess,
+			LastAck:                 lastSendAck,
 			LastNack:                lastNack,
 			LastNackMessage:         lastNackMsg,
 			LastRecvResourceSuccess: lastRecvResourceSuccess,
@@ -753,7 +751,7 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 
 		expect := Status{
 			Connected:               true,
-			LastAck:                 lastSendSuccess,
+			LastAck:                 lastSendAck,
 			LastNack:                lastNack,
 			LastNackMessage:         lastNackMsg,
 			LastRecvResourceSuccess: lastRecvResourceSuccess,
@@ -785,7 +783,7 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 
 		expect := Status{
 			Connected:               true,
-			LastAck:                 lastSendSuccess,
+			LastAck:                 lastSendAck,
 			LastNack:                lastNack,
 			LastNackMessage:         lastNackMsg,
 			LastRecvResourceSuccess: lastRecvResourceSuccess,
@@ -816,7 +814,7 @@ func TestStreamResources_Server_StreamTracker(t *testing.T) {
 		expect := Status{
 			Connected:               false,
 			DisconnectErrorMessage:  lastRecvErrorMsg,
-			LastAck:                 lastSendSuccess,
+			LastAck:                 lastSendAck,
 			LastNack:                lastNack,
 			LastNackMessage:         lastNackMsg,
 			DisconnectTime:          disconnectTime,
@@ -1128,9 +1126,9 @@ func TestStreamResources_Server_DisconnectsOnHeartbeatTimeout(t *testing.T) {
 	}
 
 	srv, store := newTestServer(t, func(c *Config) {
-		c.Tracker.SetClock(it.Now)
 		c.incomingHeartbeatTimeout = 5 * time.Millisecond
 	})
+	srv.Tracker.setClock(it.Now)
 
 	p := writePeeringToBeDialed(t, store, 1, "my-peer")
 	require.Empty(t, p.PeerID, "should be empty if being dialed")
@@ -1176,9 +1174,9 @@ func TestStreamResources_Server_SendsHeartbeats(t *testing.T) {
 	outgoingHeartbeatInterval := 5 * time.Millisecond
 
 	srv, store := newTestServer(t, func(c *Config) {
-		c.Tracker.SetClock(it.Now)
 		c.outgoingHeartbeatInterval = outgoingHeartbeatInterval
 	})
+	srv.Tracker.setClock(it.Now)
 
 	p := writePeeringToBeDialed(t, store, 1, "my-peer")
 	require.Empty(t, p.PeerID, "should be empty if being dialed")
@@ -1235,9 +1233,9 @@ func TestStreamResources_Server_KeepsConnectionOpenWithHeartbeat(t *testing.T) {
 	incomingHeartbeatTimeout := 10 * time.Millisecond
 
 	srv, store := newTestServer(t, func(c *Config) {
-		c.Tracker.SetClock(it.Now)
 		c.incomingHeartbeatTimeout = incomingHeartbeatTimeout
 	})
+	srv.Tracker.setClock(it.Now)
 
 	p := writePeeringToBeDialed(t, store, 1, "my-peer")
 	require.Empty(t, p.PeerID, "should be empty if being dialed")
@@ -2746,7 +2744,6 @@ func newTestServer(t *testing.T, configFn func(c *Config)) (*testServer, *state.
 			store: store,
 			pub:   publisher,
 		},
-		Tracker:        NewTracker(),
 		GetStore:       func() StateStore { return store },
 		Logger:         testutil.Logger(t),
 		Datacenter:     "dc1",

agent/grpc-external/services/peerstream/stream_tracker.go

@@ -14,18 +14,27 @@ type Tracker struct {
 	mu      sync.RWMutex
 	streams map[string]*MutableStatus
 
+	// heartbeatTimeout is the max duration a connection is allowed to be
+	// disconnected before the stream health is reported as non-healthy
+	heartbeatTimeout time.Duration
+
 	// timeNow is a shim for testing.
 	timeNow func() time.Time
 }
 
-func NewTracker() *Tracker {
+func NewTracker(heartbeatTimeout time.Duration) *Tracker {
+	if heartbeatTimeout == 0 {
+		heartbeatTimeout = defaultIncomingHeartbeatTimeout
+	}
 	return &Tracker{
-		streams: make(map[string]*MutableStatus),
+		streams:          make(map[string]*MutableStatus),
-		timeNow: time.Now,
+		timeNow:          time.Now,
+		heartbeatTimeout: heartbeatTimeout,
 	}
 }
 
-func (t *Tracker) SetClock(clock func() time.Time) {
+// setClock is used for debugging purposes only.
+func (t *Tracker) setClock(clock func() time.Time) {
 	if clock == nil {
 		t.timeNow = time.Now
 	} else {
@@ -101,7 +110,9 @@ func (t *Tracker) StreamStatus(id string) (resp Status, found bool) {
 
 	s, ok := t.streams[id]
 	if !ok {
-		return Status{}, false
+		return Status{
+			NeverConnected: true,
+		}, false
 	}
 	return s.GetStatus(), true
 }
@@ -126,6 +137,39 @@ func (t *Tracker) DeleteStatus(id string) {
 	delete(t.streams, id)
 }
 
+// IsHealthy is a calculates the health of a peering status.
+// We define a peering as unhealthy if its status has been in the following
+// states for longer than the configured incomingHeartbeatTimeout.
+//   - If it is disconnected
+//   - If the last received Nack is newer than last received Ack
+//   - If the last received error is newer than last received success
+//
+// If none of these conditions apply, we call the peering healthy.
+func (t *Tracker) IsHealthy(s Status) bool {
+	// If stream is in a disconnected state for longer than the configured
+	// heartbeat timeout, report as unhealthy.
+	if !s.DisconnectTime.IsZero() &&
+		t.timeNow().Sub(s.DisconnectTime) > t.heartbeatTimeout {
+		return false
+	}
+
+	// If last Nack is after last Ack, it means the peer is unable to
+	// handle our replication message.
+	if s.LastNack.After(s.LastAck) &&
+		t.timeNow().Sub(s.LastAck) > t.heartbeatTimeout {
+		return false
+	}
+
+	// If last recv error is newer than last recv success, we were unable
+	// to handle the peer's replication message.
+	if s.LastRecvError.After(s.LastRecvResourceSuccess) &&
+		t.timeNow().Sub(s.LastRecvError) > t.heartbeatTimeout {
+		return false
+	}
+
+	return true
+}
+
 type MutableStatus struct {
 	mu sync.RWMutex
 
@@ -145,6 +189,9 @@ type Status struct {
 	// Connected is true when there is an open stream for the peer.
 	Connected bool
 
+	// NeverConnected is true for peerings that have never connected, false otherwise.
+	NeverConnected bool
+
 	// DisconnectErrorMessage tracks the error that caused the stream to disconnect non-gracefully.
 	// If the stream is connected or it disconnected gracefully it will be empty.
 	DisconnectErrorMessage string
@@ -199,7 +246,8 @@ func (s *Status) GetExportedServicesCount() uint64 {
 func newMutableStatus(now func() time.Time, connected bool) *MutableStatus {
 	return &MutableStatus{
 		Status: Status{
-			Connected: connected,
+			Connected:      connected,
+			NeverConnected: !connected,
 		},
 		timeNow: now,
 		doneCh:  make(chan struct{}),

agent/grpc-external/services/peerstream/stream_tracker_test.go

@@ -5,13 +5,117 @@ import (
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
 	"github.com/hashicorp/consul/sdk/testutil"
 )
 
+const (
+	aPeerID = "63b60245-c475-426b-b314-4588d210859d"
+)
+
+func TestTracker_IsHealthy(t *testing.T) {
+	type testcase struct {
+		name         string
+		tracker      *Tracker
+		modifierFunc func(status *MutableStatus)
+		expectedVal  bool
+	}
+
+	tcs := []testcase{
+		{
+			name:        "disconnect time within timeout",
+			tracker:     NewTracker(defaultIncomingHeartbeatTimeout),
+			expectedVal: true,
+			modifierFunc: func(status *MutableStatus) {
+				status.DisconnectTime = time.Now()
+			},
+		},
+		{
+			name:        "disconnect time past timeout",
+			tracker:     NewTracker(1 * time.Millisecond),
+			expectedVal: false,
+			modifierFunc: func(status *MutableStatus) {
+				status.DisconnectTime = time.Now().Add(-1 * time.Minute)
+			},
+		},
+		{
+			name:        "receive error before receive success within timeout",
+			tracker:     NewTracker(defaultIncomingHeartbeatTimeout),
+			expectedVal: true,
+			modifierFunc: func(status *MutableStatus) {
+				now := time.Now()
+				status.LastRecvResourceSuccess = now
+				status.LastRecvError = now.Add(1 * time.Second)
+			},
+		},
+		{
+			name:        "receive error before receive success within timeout",
+			tracker:     NewTracker(defaultIncomingHeartbeatTimeout),
+			expectedVal: true,
+			modifierFunc: func(status *MutableStatus) {
+				now := time.Now()
+				status.LastRecvResourceSuccess = now
+				status.LastRecvError = now.Add(1 * time.Second)
+			},
+		},
+		{
+			name:        "receive error before receive success past timeout",
+			tracker:     NewTracker(1 * time.Millisecond),
+			expectedVal: false,
+			modifierFunc: func(status *MutableStatus) {
+				now := time.Now().Add(-2 * time.Second)
+				status.LastRecvResourceSuccess = now
+				status.LastRecvError = now.Add(1 * time.Second)
+			},
+		},
+		{
+			name:        "nack before ack within timeout",
+			tracker:     NewTracker(defaultIncomingHeartbeatTimeout),
+			expectedVal: true,
+			modifierFunc: func(status *MutableStatus) {
+				now := time.Now()
+				status.LastAck = now
+				status.LastNack = now.Add(1 * time.Second)
+			},
+		},
+		{
+			name:        "nack before ack past timeout",
+			tracker:     NewTracker(1 * time.Millisecond),
+			expectedVal: false,
+			modifierFunc: func(status *MutableStatus) {
+				now := time.Now().Add(-2 * time.Second)
+				status.LastAck = now
+				status.LastNack = now.Add(1 * time.Second)
+			},
+		},
+		{
+			name:        "healthy",
+			tracker:     NewTracker(defaultIncomingHeartbeatTimeout),
+			expectedVal: true,
+		},
+	}
+
+	for _, tc := range tcs {
+		t.Run(tc.name, func(t *testing.T) {
+			tracker := tc.tracker
+
+			st, err := tracker.Connected(aPeerID)
+			require.NoError(t, err)
+			require.True(t, st.Connected)
+
+			if tc.modifierFunc != nil {
+				tc.modifierFunc(st)
+			}
+
+			assert.Equal(t, tc.expectedVal, tracker.IsHealthy(st.GetStatus()))
+		})
+	}
+}
+
 func TestTracker_EnsureConnectedDisconnected(t *testing.T) {
-	tracker := NewTracker()
+	tracker := NewTracker(defaultIncomingHeartbeatTimeout)
 	peerID := "63b60245-c475-426b-b314-4588d210859d"
 
 	it := incrementalTime{
@@ -96,7 +200,7 @@ func TestTracker_EnsureConnectedDisconnected(t *testing.T) {
 
 		status, ok := tracker.StreamStatus(peerID)
 		require.False(t, ok)
-		require.Zero(t, status)
+		require.Equal(t, Status{NeverConnected: true}, status)
 	})
 }
 
@@ -108,7 +212,7 @@ func TestTracker_connectedStreams(t *testing.T) {
 	}
 
 	run := func(t *testing.T, tc testCase) {
-		tracker := NewTracker()
+		tracker := NewTracker(defaultIncomingHeartbeatTimeout)
 		if tc.setup != nil {
 			tc.setup(t, tracker)
 		}

agent/proxycfg-glue/glue.go

@@ -124,15 +124,21 @@ func (c *cacheProxyDataSource[ReqType]) Notify(
 
 func dispatchCacheUpdate(ch chan<- proxycfg.UpdateEvent) cache.Callback {
 	return func(ctx context.Context, e cache.UpdateEvent) {
-		u := proxycfg.UpdateEvent{
-			CorrelationID: e.CorrelationID,
-			Result:        e.Result,
-			Err:           e.Err,
-		}
-
 		select {
-		case ch <- u:
+		case ch <- newUpdateEvent(e.CorrelationID, e.Result, e.Err):
 		case <-ctx.Done():
 		}
 	}
 }
+
+func newUpdateEvent(correlationID string, result any, err error) proxycfg.UpdateEvent {
+	// This roughly matches the logic in agent/submatview.LocalMaterializer.isTerminalError.
+	if acl.IsErrNotFound(err) {
+		err = proxycfg.TerminalError(err)
+	}
+	return proxycfg.UpdateEvent{
+		CorrelationID: correlationID,
+		Result:        result,
+		Err:           err,
+	}
+}

agent/proxycfg-glue/intention_upstreams.go

@@ -54,13 +54,8 @@ func (s serverIntentionUpstreams) Notify(ctx context.Context, req *structs.Servi
 
 func dispatchBlockingQueryUpdate[ResultType any](ch chan<- proxycfg.UpdateEvent) func(context.Context, string, ResultType, error) {
 	return func(ctx context.Context, correlationID string, result ResultType, err error) {
-		event := proxycfg.UpdateEvent{
-			CorrelationID: correlationID,
-			Result:        result,
-			Err:           err,
-		}
 		select {
-		case ch <- event:
+		case ch <- newUpdateEvent(correlationID, result, err):
 		case <-ctx.Done():
 		}
 	}

agent/proxycfg-glue/intentions.go

@@ -39,12 +39,8 @@ func (c cacheIntentions) Notify(ctx context.Context, req *structs.ServiceSpecifi
 		QueryOptions: structs.QueryOptions{Token: req.QueryOptions.Token},
 	}
 	return c.c.NotifyCallback(ctx, cachetype.IntentionMatchName, query, correlationID, func(ctx context.Context, event cache.UpdateEvent) {
-		e := proxycfg.UpdateEvent{
+		var result any
-			CorrelationID: correlationID,
+		if event.Err == nil {
-			Err:           event.Err,
-		}
-
-		if e.Err == nil {
 			rsp, ok := event.Result.(*structs.IndexedIntentionMatches)
 			if !ok {
 				return
@@ -54,11 +50,11 @@ func (c cacheIntentions) Notify(ctx context.Context, req *structs.ServiceSpecifi
 			if len(rsp.Matches) != 0 {
 				matches = rsp.Matches[0]
 			}
-			e.Result = matches
+			result = matches
 		}
 
 		select {
-		case ch <- e:
+		case ch <- newUpdateEvent(correlationID, result, event.Err):
 		case <-ctx.Done():
 		}
 	})
@@ -110,10 +106,7 @@ func (s *serverIntentions) Notify(ctx context.Context, req *structs.ServiceSpeci
 
 		sort.Sort(structs.IntentionPrecedenceSorter(intentions))
 
-		return proxycfg.UpdateEvent{
+		return newUpdateEvent(correlationID, intentions, nil), true
-			CorrelationID: correlationID,
-			Result:        intentions,
-		}, true
 	}
 
 	for subjectIdx, subject := range subjects {

agent/proxycfg/connect_proxy.go

@@ -280,16 +280,6 @@ func (s *handlerConnectProxy) handleUpdate(ctx context.Context, u UpdateEvent, s
 		}
 		snap.Roots = roots
 
-	case strings.HasPrefix(u.CorrelationID, peerTrustBundleIDPrefix):
-		resp, ok := u.Result.(*pbpeering.TrustBundleReadResponse)
-		if !ok {
-			return fmt.Errorf("invalid type for response: %T", u.Result)
-		}
-		peer := strings.TrimPrefix(u.CorrelationID, peerTrustBundleIDPrefix)
-		if resp.Bundle != nil {
-			snap.ConnectProxy.UpstreamPeerTrustBundles.Set(peer, resp.Bundle)
-		}
-
 	case u.CorrelationID == peeringTrustBundlesWatchID:
 		resp, ok := u.Result.(*pbpeering.TrustBundleListByServiceResponse)
 		if !ok {
@@ -369,6 +359,17 @@ func (s *handlerConnectProxy) handleUpdate(ctx context.Context, u UpdateEvent, s
 		// Clean up data
 		//
 
+		peeredChainTargets := make(map[UpstreamID]struct{})
+		for _, discoChain := range snap.ConnectProxy.DiscoveryChain {
+			for _, target := range discoChain.Targets {
+				if target.Peer == "" {
+					continue
+				}
+				uid := NewUpstreamIDFromTargetID(target.ID)
+				peeredChainTargets[uid] = struct{}{}
+			}
+		}
+
 		validPeerNames := make(map[string]struct{})
 
 		// Iterate through all known endpoints and remove references to upstream IDs that weren't in the update
@@ -383,6 +384,11 @@ func (s *handlerConnectProxy) handleUpdate(ctx context.Context, u UpdateEvent, s
 				validPeerNames[uid.Peer] = struct{}{}
 				return true
 			}
+			// Peered upstream came from a discovery chain target
+			if _, ok := peeredChainTargets[uid]; ok {
+				validPeerNames[uid.Peer] = struct{}{}
+				return true
+			}
 			snap.ConnectProxy.PeerUpstreamEndpoints.CancelWatch(uid)
 			return true
 		})
@@ -463,8 +469,14 @@ func (s *handlerConnectProxy) handleUpdate(ctx context.Context, u UpdateEvent, s
 				continue
 			}
 			if _, ok := seenUpstreams[uid]; !ok {
-				for _, cancelFn := range targets {
+				for targetID, cancelFn := range targets {
 					cancelFn()
+
+					targetUID := NewUpstreamIDFromTargetID(targetID)
+					if targetUID.Peer != "" {
+						snap.ConnectProxy.PeerUpstreamEndpoints.CancelWatch(targetUID)
+						snap.ConnectProxy.UpstreamPeerTrustBundles.CancelWatch(targetUID.Peer)
+					}
 				}
 				delete(snap.ConnectProxy.WatchedUpstreams, uid)
 			}

agent/proxycfg/data_sources.go

@@ -2,6 +2,7 @@ package proxycfg
 
 import (
 	"context"
+	"errors"
 
 	cachetype "github.com/hashicorp/consul/agent/cache-types"
 	"github.com/hashicorp/consul/agent/structs"
@@ -15,6 +16,28 @@ type UpdateEvent struct {
 	Err           error
 }
 
+// TerminalError wraps the given error to indicate that the data source is in
+// an irrecoverably broken state (e.g. because the given ACL token has been
+// deleted).
+//
+// Setting UpdateEvent.Err to a TerminalError causes all watches to be canceled
+// which, in turn, terminates the xDS streams.
+func TerminalError(err error) error {
+	return terminalError{err}
+}
+
+// IsTerminalError returns whether the given error indicates that the data
+// source is in an irrecoverably broken state so watches should be torn down
+// and retried at a higher level.
+func IsTerminalError(err error) bool {
+	return errors.As(err, &terminalError{})
+}
+
+type terminalError struct{ err error }
+
+func (e terminalError) Error() string { return e.err.Error() }
+func (e terminalError) Unwrap() error { return e.err }
+
 // DataSources contains the dependencies used to consume data used to configure
 // proxies.
 type DataSources struct {

agent/proxycfg/ingress_gateway.go

@@ -5,7 +5,9 @@ import (
 	"fmt"
 
 	cachetype "github.com/hashicorp/consul/agent/cache-types"
+	"github.com/hashicorp/consul/agent/proxycfg/internal/watch"
 	"github.com/hashicorp/consul/agent/structs"
+	"github.com/hashicorp/consul/proto/pbpeering"
 )
 
 type handlerIngressGateway struct {
@@ -66,6 +68,9 @@ func (s *handlerIngressGateway) initialize(ctx context.Context) (ConfigSnapshot,
 	snap.IngressGateway.WatchedGateways = make(map[UpstreamID]map[string]context.CancelFunc)
 	snap.IngressGateway.WatchedGatewayEndpoints = make(map[UpstreamID]map[string]structs.CheckServiceNodes)
 	snap.IngressGateway.Listeners = make(map[IngressListenerKey]structs.IngressListener)
+	snap.IngressGateway.UpstreamPeerTrustBundles = watch.NewMap[string, *pbpeering.PeeringTrustBundle]()
+	snap.IngressGateway.PeerUpstreamEndpoints = watch.NewMap[UpstreamID, structs.CheckServiceNodes]()
+	snap.IngressGateway.PeerUpstreamEndpointsUseHostnames = make(map[UpstreamID]struct{})
 	return snap, nil
 }
 
@@ -152,6 +157,12 @@ func (s *handlerIngressGateway) handleUpdate(ctx context.Context, u UpdateEvent,
 					delete(snap.IngressGateway.WatchedUpstreams[uid], targetID)
 					delete(snap.IngressGateway.WatchedUpstreamEndpoints[uid], targetID)
 					cancelUpstreamFn()
+
+					targetUID := NewUpstreamIDFromTargetID(targetID)
+					if targetUID.Peer != "" {
+						snap.IngressGateway.PeerUpstreamEndpoints.CancelWatch(targetUID)
+						snap.IngressGateway.UpstreamPeerTrustBundles.CancelWatch(targetUID.Peer)
+					}
 				}
 
 				cancelFn()

agent/proxycfg/manager.go

@@ -127,7 +127,7 @@ func (m *Manager) Register(id ProxyID, ns *structs.NodeService, source ProxySour
 		}
 
 		// We are updating the proxy, close its old state
-		state.Close()
+		state.Close(false)
 	}
 
 	// TODO: move to a function that translates ManagerConfig->stateConfig
@@ -148,14 +148,13 @@ func (m *Manager) Register(id ProxyID, ns *structs.NodeService, source ProxySour
 		return err
 	}
 
-	ch, err := state.Watch()
+	if _, err = state.Watch(); err != nil {
-	if err != nil {
 		return err
 	}
 	m.proxies[id] = state
 
 	// Start a goroutine that will wait for changes and broadcast them to watchers.
-	go m.notifyBroadcast(ch)
+	go m.notifyBroadcast(id, state)
 	return nil
 }
 
@@ -175,8 +174,8 @@ func (m *Manager) Deregister(id ProxyID, source ProxySource) {
 	}
 
 	// Closing state will let the goroutine we started in Register finish since
-	// watch chan is closed.
+	// watch chan is closed
-	state.Close()
+	state.Close(false)
 	delete(m.proxies, id)
 
 	// We intentionally leave potential watchers hanging here - there is no new
@@ -186,11 +185,17 @@ func (m *Manager) Deregister(id ProxyID, source ProxySource) {
 	// cleaned up naturally.
 }
 
-func (m *Manager) notifyBroadcast(ch <-chan ConfigSnapshot) {
+func (m *Manager) notifyBroadcast(proxyID ProxyID, state *state) {
-	// Run until ch is closed
+	// Run until ch is closed (by a defer in state.run).
-	for snap := range ch {
+	for snap := range state.snapCh {
 		m.notify(&snap)
 	}
+
+	// If state.run exited because of an irrecoverable error, close all of the
+	// watchers so that the consumers reconnect/retry at a higher level.
+	if state.failed() {
+		m.closeAllWatchers(proxyID)
+	}
 }
 
 func (m *Manager) notify(snap *ConfigSnapshot) {
@@ -281,6 +286,20 @@ func (m *Manager) Watch(id ProxyID) (<-chan *ConfigSnapshot, CancelFunc) {
 	}
 }
 
+func (m *Manager) closeAllWatchers(proxyID ProxyID) {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+
+	watchers, ok := m.watchers[proxyID]
+	if !ok {
+		return
+	}
+
+	for watchID := range watchers {
+		m.closeWatchLocked(proxyID, watchID)
+	}
+}
+
 // closeWatchLocked cleans up state related to a single watcher. It assumes the
 // lock is held.
 func (m *Manager) closeWatchLocked(proxyID ProxyID, watchID uint64) {
@@ -309,7 +328,7 @@ func (m *Manager) Close() error {
 
 	// Then close all states
 	for proxyID, state := range m.proxies {
-		state.Close()
+		state.Close(false)
 		delete(m.proxies, proxyID)
 	}
 	return nil

agent/proxycfg/naming.go

@@ -63,22 +63,29 @@ func NewUpstreamIDFromServiceID(sid structs.ServiceID) UpstreamID {
 	return id
 }
 
-// TODO(peering): confirm we don't need peername here
 func NewUpstreamIDFromTargetID(tid string) UpstreamID {
-	// Drop the leading subset if one is present in the target ID.
+	var id UpstreamID
-	separators := strings.Count(tid, ".")
+	split := strings.Split(tid, ".")
-	if separators > 3 {
+
-		prefix := tid[:strings.Index(tid, ".")+1]
+	switch {
-		tid = strings.TrimPrefix(tid, prefix)
+	case split[len(split)-2] == "external":
+		id = UpstreamID{
+			Name:           split[0],
+			EnterpriseMeta: acl.NewEnterpriseMetaWithPartition(split[2], split[1]),
+			Peer:           split[4],
+		}
+	case len(split) == 5:
+		// Drop the leading subset if one is present in the target ID.
+		split = split[1:]
+		fallthrough
+	default:
+		id = UpstreamID{
+			Name:           split[0],
+			EnterpriseMeta: acl.NewEnterpriseMetaWithPartition(split[2], split[1]),
+			Datacenter:     split[3],
+		}
 	}
 
-	split := strings.SplitN(tid, ".", 4)
-
-	id := UpstreamID{
-		Name:           split[0],
-		EnterpriseMeta: acl.NewEnterpriseMetaWithPartition(split[2], split[1]),
-		Datacenter:     split[3],
-	}
 	id.normalize()
 	return id
 }

agent/proxycfg/naming_test.go

@@ -35,6 +35,13 @@ func TestUpstreamIDFromTargetID(t *testing.T) {
 				Datacenter: "dc2",
 			},
 		},
+		"peered": {
+			tid: "foo.default.default.external.cluster-01",
+			expect: UpstreamID{
+				Name: "foo",
+				Peer: "cluster-01",
+			},
+		},
 	}
 
 	for name, tc := range cases {

agent/proxycfg/snapshot.go

@@ -814,6 +814,18 @@ func (s *ConfigSnapshot) MeshConfigTLSOutgoing() *structs.MeshDirectionalTLSConf
 	return mesh.TLS.Outgoing
 }
 
+func (s *ConfigSnapshot) ToConfigSnapshotUpstreams() (*ConfigSnapshotUpstreams, error) {
+	switch s.Kind {
+	case structs.ServiceKindConnectProxy:
+		return &s.ConnectProxy.ConfigSnapshotUpstreams, nil
+	case structs.ServiceKindIngressGateway:
+		return &s.IngressGateway.ConfigSnapshotUpstreams, nil
+	default:
+		// This is a coherence check and should never fail
+		return nil, fmt.Errorf("No upstream snapshot for gateway mode %q", s.Kind)
+	}
+}
+
 func (u *ConfigSnapshotUpstreams) UpstreamPeerMeta(uid UpstreamID) structs.PeeringServiceMeta {
 	nodes, _ := u.PeerUpstreamEndpoints.Get(uid)
 	if len(nodes) == 0 {

agent/proxycfg/state.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"reflect"
+	"sync/atomic"
 	"time"
 
 	"github.com/hashicorp/go-hclog"
@@ -70,11 +71,21 @@ type state struct {
 	// in Watch.
 	cancel func()
 
+	// failedFlag is (atomically) set to 1 (by Close) when run exits because a data
+	// source is in an irrecoverable state. It can be read with failed.
+	failedFlag int32
+
 	ch     chan UpdateEvent
 	snapCh chan ConfigSnapshot
 	reqCh  chan chan *ConfigSnapshot
 }
 
+// failed returns whether run exited because a data source is in an
+// irrecoverable state.
+func (s *state) failed() bool {
+	return atomic.LoadInt32(&s.failedFlag) == 1
+}
+
 type DNSConfig struct {
 	Domain    string
 	AltDomain string
@@ -250,10 +261,13 @@ func (s *state) Watch() (<-chan ConfigSnapshot, error) {
 }
 
 // Close discards the state and stops any long-running watches.
-func (s *state) Close() error {
+func (s *state) Close(failed bool) error {
 	if s.cancel != nil {
 		s.cancel()
 	}
+	if failed {
+		atomic.StoreInt32(&s.failedFlag, 1)
+	}
 	return nil
 }
 
@@ -300,7 +314,13 @@ func (s *state) run(ctx context.Context, snap *ConfigSnapshot) {
 		case <-ctx.Done():
 			return
 		case u := <-s.ch:
-			s.logger.Trace("A blocking query returned; handling snapshot update", "correlationID", u.CorrelationID)
+			s.logger.Trace("Data source returned; handling snapshot update", "correlationID", u.CorrelationID)
+
+			if IsTerminalError(u.Err) {
+				s.logger.Error("Data source in an irrecoverable state; exiting", "error", u.Err, "correlationID", u.CorrelationID)
+				s.Close(true)
+				return
+			}
 
 			if err := s.handler.handleUpdate(ctx, u, snap); err != nil {
 				s.logger.Error("Failed to handle update from watch",

agent/proxycfg/state_test.go

@@ -493,6 +493,11 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 							Mode: structs.MeshGatewayModeNone,
 						},
 					},
+					structs.Upstream{
+						DestinationType: structs.UpstreamDestTypeService,
+						DestinationName: "api-failover-to-peer",
+						LocalBindPort:   10007,
+					},
 					structs.Upstream{
 						DestinationType: structs.UpstreamDestTypeService,
 						DestinationName: "api-dc2",
@@ -552,6 +557,16 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 						Mode: structs.MeshGatewayModeNone,
 					},
 				}),
+				fmt.Sprintf("discovery-chain:%s-failover-to-peer", apiUID.String()): genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
+					Name:                 "api-failover-to-peer",
+					EvaluateInDatacenter: "dc1",
+					EvaluateInNamespace:  "default",
+					EvaluateInPartition:  "default",
+					Datacenter:           "dc1",
+					OverrideMeshGateway: structs.MeshGatewayConfig{
+						Mode: meshGatewayProxyConfigValue,
+					},
+				}),
 				fmt.Sprintf("discovery-chain:%s-dc2", apiUID.String()): genVerifyDiscoveryChainWatch(&structs.DiscoveryChainRequest{
 					Name:                 "api-dc2",
 					EvaluateInDatacenter: "dc1",
@@ -639,6 +654,26 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 					},
 					Err: nil,
 				},
+				{
+					CorrelationID: fmt.Sprintf("discovery-chain:%s-failover-to-peer", apiUID.String()),
+					Result: &structs.DiscoveryChainResponse{
+						Chain: discoverychain.TestCompileConfigEntries(t, "api-failover-to-peer", "default", "default", "dc1", "trustdomain.consul",
+							func(req *discoverychain.CompileRequest) {
+								req.OverrideMeshGateway.Mode = meshGatewayProxyConfigValue
+							}, &structs.ServiceResolverConfigEntry{
+								Kind: structs.ServiceResolver,
+								Name: "api-failover-to-peer",
+								Failover: map[string]structs.ServiceResolverFailover{
+									"*": {
+										Targets: []structs.ServiceResolverFailoverTarget{
+											{Peer: "cluster-01"},
+										},
+									},
+								},
+							}),
+					},
+					Err: nil,
+				},
 			},
 			verifySnapshot: func(t testing.TB, snap *ConfigSnapshot) {
 				require.True(t, snap.Valid())
@@ -646,15 +681,18 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 				require.Equal(t, indexedRoots, snap.Roots)
 
 				require.Equal(t, issuedCert, snap.ConnectProxy.Leaf)
-				require.Len(t, snap.ConnectProxy.DiscoveryChain, 5, "%+v", snap.ConnectProxy.DiscoveryChain)
+				require.Len(t, snap.ConnectProxy.DiscoveryChain, 6, "%+v", snap.ConnectProxy.DiscoveryChain)
-				require.Len(t, snap.ConnectProxy.WatchedUpstreams, 5, "%+v", snap.ConnectProxy.WatchedUpstreams)
+				require.Len(t, snap.ConnectProxy.WatchedUpstreams, 6, "%+v", snap.ConnectProxy.WatchedUpstreams)
-				require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 5, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
+				require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 6, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
-				require.Len(t, snap.ConnectProxy.WatchedGateways, 5, "%+v", snap.ConnectProxy.WatchedGateways)
+				require.Len(t, snap.ConnectProxy.WatchedGateways, 6, "%+v", snap.ConnectProxy.WatchedGateways)
-				require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 5, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
+				require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 6, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
 
 				require.Len(t, snap.ConnectProxy.WatchedServiceChecks, 0, "%+v", snap.ConnectProxy.WatchedServiceChecks)
 				require.Len(t, snap.ConnectProxy.PreparedQueryEndpoints, 0, "%+v", snap.ConnectProxy.PreparedQueryEndpoints)
 
+				require.Equal(t, 1, snap.ConnectProxy.ConfigSnapshotUpstreams.PeerUpstreamEndpoints.Len())
+				require.Equal(t, 1, snap.ConnectProxy.ConfigSnapshotUpstreams.UpstreamPeerTrustBundles.Len())
+
 				require.True(t, snap.ConnectProxy.IntentionsSet)
 				require.Equal(t, ixnMatch, snap.ConnectProxy.Intentions)
 				require.True(t, snap.ConnectProxy.MeshConfigSet)
@@ -667,6 +705,7 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 				fmt.Sprintf("upstream-target:api-failover-remote.default.default.dc2:%s-failover-remote?dc=dc2", apiUID.String()): genVerifyServiceSpecificRequest("api-failover-remote", "", "dc2", true),
 				fmt.Sprintf("upstream-target:api-failover-local.default.default.dc2:%s-failover-local?dc=dc2", apiUID.String()):   genVerifyServiceSpecificRequest("api-failover-local", "", "dc2", true),
 				fmt.Sprintf("upstream-target:api-failover-direct.default.default.dc2:%s-failover-direct?dc=dc2", apiUID.String()): genVerifyServiceSpecificRequest("api-failover-direct", "", "dc2", true),
+				upstreamPeerWatchIDPrefix + fmt.Sprintf("%s-failover-to-peer?peer=cluster-01", apiUID.String()):                   genVerifyServiceSpecificPeeredRequest("api-failover-to-peer", "", "", "cluster-01", true),
 				fmt.Sprintf("mesh-gateway:dc2:%s-failover-remote?dc=dc2", apiUID.String()):                                        genVerifyGatewayWatch("dc2"),
 				fmt.Sprintf("mesh-gateway:dc1:%s-failover-local?dc=dc2", apiUID.String()):                                         genVerifyGatewayWatch("dc1"),
 			},
@@ -676,15 +715,18 @@ func TestState_WatchesAndUpdates(t *testing.T) {
 				require.Equal(t, indexedRoots, snap.Roots)
 
 				require.Equal(t, issuedCert, snap.ConnectProxy.Leaf)
-				require.Len(t, snap.ConnectProxy.DiscoveryChain, 5, "%+v", snap.ConnectProxy.DiscoveryChain)
+				require.Len(t, snap.ConnectProxy.DiscoveryChain, 6, "%+v", snap.ConnectProxy.DiscoveryChain)
-				require.Len(t, snap.ConnectProxy.WatchedUpstreams, 5, "%+v", snap.ConnectProxy.WatchedUpstreams)
+				require.Len(t, snap.ConnectProxy.WatchedUpstreams, 6, "%+v", snap.ConnectProxy.WatchedUpstreams)
-				require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 5, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
+				require.Len(t, snap.ConnectProxy.WatchedUpstreamEndpoints, 6, "%+v", snap.ConnectProxy.WatchedUpstreamEndpoints)
-				require.Len(t, snap.ConnectProxy.WatchedGateways, 5, "%+v", snap.ConnectProxy.WatchedGateways)
+				require.Len(t, snap.ConnectProxy.WatchedGateways, 6, "%+v", snap.ConnectProxy.WatchedGateways)
-				require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 5, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
+				require.Len(t, snap.ConnectProxy.WatchedGatewayEndpoints, 6, "%+v", snap.ConnectProxy.WatchedGatewayEndpoints)
 
 				require.Len(t, snap.ConnectProxy.WatchedServiceChecks, 0, "%+v", snap.ConnectProxy.WatchedServiceChecks)
 				require.Len(t, snap.ConnectProxy.PreparedQueryEndpoints, 0, "%+v", snap.ConnectProxy.PreparedQueryEndpoints)
 
+				require.Equal(t, 1, snap.ConnectProxy.ConfigSnapshotUpstreams.PeerUpstreamEndpoints.Len())
+				require.Equal(t, 1, snap.ConnectProxy.ConfigSnapshotUpstreams.UpstreamPeerTrustBundles.Len())
+
 				require.True(t, snap.ConnectProxy.IntentionsSet)
 				require.Equal(t, ixnMatch, snap.ConnectProxy.Intentions)
 			},

agent/proxycfg/testing.go

@@ -280,6 +280,31 @@ func TestUpstreamNodesDC2(t testing.T) structs.CheckServiceNodes {
 	}
 }
 
+func TestUpstreamNodesPeerCluster01(t testing.T) structs.CheckServiceNodes {
+	peer := "cluster-01"
+	service := structs.TestNodeServiceWithNameInPeer(t, "web", peer)
+	return structs.CheckServiceNodes{
+		structs.CheckServiceNode{
+			Node: &structs.Node{
+				ID:       "test1",
+				Node:     "test1",
+				Address:  "10.40.1.1",
+				PeerName: peer,
+			},
+			Service: service,
+		},
+		structs.CheckServiceNode{
+			Node: &structs.Node{
+				ID:       "test2",
+				Node:     "test2",
+				Address:  "10.40.1.2",
+				PeerName: peer,
+			},
+			Service: service,
+		},
+	}
+}
+
 func TestUpstreamNodesInStatusDC2(t testing.T, status string) structs.CheckServiceNodes {
 	return structs.CheckServiceNodes{
 		structs.CheckServiceNode{

agent/proxycfg/testing_upstreams.go

@@ -8,6 +8,7 @@ import (
 	"github.com/hashicorp/consul/agent/connect"
 	"github.com/hashicorp/consul/agent/consul/discoverychain"
 	"github.com/hashicorp/consul/agent/structs"
+	"github.com/hashicorp/consul/proto/pbpeering"
 )
 
 func setupTestVariationConfigEntriesAndSnapshot(
@@ -72,6 +73,24 @@ func setupTestVariationConfigEntriesAndSnapshot(
 				Nodes: TestGatewayNodesDC2(t),
 			},
 		})
+	case "failover-to-cluster-peer":
+		events = append(events, UpdateEvent{
+			CorrelationID: "peer-trust-bundle:cluster-01",
+			Result: &pbpeering.TrustBundleReadResponse{
+				Bundle: &pbpeering.PeeringTrustBundle{
+					PeerName:          "peer1",
+					TrustDomain:       "peer1.domain",
+					ExportedPartition: "peer1ap",
+					RootPEMs:          []string{"peer1-root-1"},
+				},
+			},
+		})
+		events = append(events, UpdateEvent{
+			CorrelationID: "upstream-peer:db?peer=cluster-01",
+			Result: &structs.IndexedCheckServiceNodes{
+				Nodes: TestUpstreamNodesPeerCluster01(t),
+			},
+		})
 	case "failover-through-double-remote-gateway-triggered":
 		events = append(events, UpdateEvent{
 			CorrelationID: "upstream-target:db.default.default.dc1:" + dbUID.String(),
@@ -255,6 +274,21 @@ func setupTestVariationDiscoveryChain(
 				},
 			},
 		)
+	case "failover-to-cluster-peer":
+		entries = append(entries,
+			&structs.ServiceResolverConfigEntry{
+				Kind:           structs.ServiceResolver,
+				Name:           "db",
+				ConnectTimeout: 33 * time.Second,
+				Failover: map[string]structs.ServiceResolverFailover{
+					"*": {
+						Targets: []structs.ServiceResolverFailoverTarget{
+							{Peer: "cluster-01"},
+						},
+					},
+				},
+			},
+		)
 	case "failover-through-double-remote-gateway-triggered":
 		fallthrough
 	case "failover-through-double-remote-gateway":

agent/proxycfg/upstreams.go

@@ -9,7 +9,9 @@ import (
 	"github.com/mitchellh/mapstructure"
 
 	"github.com/hashicorp/consul/acl"
+	cachetype "github.com/hashicorp/consul/agent/cache-types"
 	"github.com/hashicorp/consul/agent/structs"
+	"github.com/hashicorp/consul/proto/pbpeering"
 )
 
 type handlerUpstreams struct {
@@ -21,9 +23,10 @@ func (s *handlerUpstreams) handleUpdateUpstreams(ctx context.Context, u UpdateEv
 		return fmt.Errorf("error filling agent cache: %v", u.Err)
 	}
 
-	upstreamsSnapshot := &snap.ConnectProxy.ConfigSnapshotUpstreams
+	upstreamsSnapshot, err := snap.ToConfigSnapshotUpstreams()
-	if snap.Kind == structs.ServiceKindIngressGateway {
+
-		upstreamsSnapshot = &snap.IngressGateway.ConfigSnapshotUpstreams
+	if err != nil {
+		return err
 	}
 
 	switch {
@@ -98,19 +101,16 @@ func (s *handlerUpstreams) handleUpdateUpstreams(ctx context.Context, u UpdateEv
 
 		uid := UpstreamIDFromString(uidString)
 
-		filteredNodes := hostnameEndpoints(
+		s.setPeerEndpoints(upstreamsSnapshot, uid, resp.Nodes)
-			s.logger,
+
-			GatewayKey{ /*empty so it never matches*/ },
+	case strings.HasPrefix(u.CorrelationID, peerTrustBundleIDPrefix):
-			resp.Nodes,
+		resp, ok := u.Result.(*pbpeering.TrustBundleReadResponse)
-		)
+		if !ok {
-		if len(filteredNodes) > 0 {
+			return fmt.Errorf("invalid type for response: %T", u.Result)
-			if set := upstreamsSnapshot.PeerUpstreamEndpoints.Set(uid, filteredNodes); set {
+		}
-				upstreamsSnapshot.PeerUpstreamEndpointsUseHostnames[uid] = struct{}{}
+		peer := strings.TrimPrefix(u.CorrelationID, peerTrustBundleIDPrefix)
-			}
+		if resp.Bundle != nil {
-		} else {
+			upstreamsSnapshot.UpstreamPeerTrustBundles.Set(peer, resp.Bundle)
-			if set := upstreamsSnapshot.PeerUpstreamEndpoints.Set(uid, resp.Nodes); set {
-				delete(upstreamsSnapshot.PeerUpstreamEndpointsUseHostnames, uid)
-			}
 		}
 
 	case strings.HasPrefix(u.CorrelationID, "upstream-target:"):
@@ -216,6 +216,23 @@ func removeColonPrefix(s string) (string, string, bool) {
 	return s[0:idx], s[idx+1:], true
 }
 
+func (s *handlerUpstreams) setPeerEndpoints(upstreamsSnapshot *ConfigSnapshotUpstreams, uid UpstreamID, nodes structs.CheckServiceNodes) {
+	filteredNodes := hostnameEndpoints(
+		s.logger,
+		GatewayKey{ /*empty so it never matches*/ },
+		nodes,
+	)
+	if len(filteredNodes) > 0 {
+		if set := upstreamsSnapshot.PeerUpstreamEndpoints.Set(uid, filteredNodes); set {
+			upstreamsSnapshot.PeerUpstreamEndpointsUseHostnames[uid] = struct{}{}
+		}
+	} else {
+		if set := upstreamsSnapshot.PeerUpstreamEndpoints.Set(uid, nodes); set {
+			delete(upstreamsSnapshot.PeerUpstreamEndpointsUseHostnames, uid)
+		}
+	}
+}
+
 func (s *handlerUpstreams) resetWatchesFromChain(
 	ctx context.Context,
 	uid UpstreamID,
@@ -255,6 +272,12 @@ func (s *handlerUpstreams) resetWatchesFromChain(
 		delete(snap.WatchedUpstreams[uid], targetID)
 		delete(snap.WatchedUpstreamEndpoints[uid], targetID)
 		cancelFn()
+
+		targetUID := NewUpstreamIDFromTargetID(targetID)
+		if targetUID.Peer != "" {
+			snap.PeerUpstreamEndpoints.CancelWatch(targetUID)
+			snap.UpstreamPeerTrustBundles.CancelWatch(targetUID.Peer)
+		}
 	}
 
 	var (
@@ -274,6 +297,7 @@ func (s *handlerUpstreams) resetWatchesFromChain(
 			service:    target.Service,
 			filter:     target.Subset.Filter,
 			datacenter: target.Datacenter,
+			peer:       target.Peer,
 			entMeta:    target.GetEnterpriseMetadata(),
 		}
 		err := s.watchUpstreamTarget(ctx, snap, opts)
@@ -384,6 +408,7 @@ type targetWatchOpts struct {
 	service    string
 	filter     string
 	datacenter string
+	peer       string
 	entMeta    *acl.EnterpriseMeta
 }
 
@@ -397,11 +422,17 @@ func (s *handlerUpstreams) watchUpstreamTarget(ctx context.Context, snap *Config
 	var finalMeta acl.EnterpriseMeta
 	finalMeta.Merge(opts.entMeta)
 
-	correlationID := "upstream-target:" + opts.chainID + ":" + opts.upstreamID.String()
+	uid := opts.upstreamID
+	correlationID := "upstream-target:" + opts.chainID + ":" + uid.String()
+
+	if opts.peer != "" {
+		uid = NewUpstreamIDFromTargetID(opts.chainID)
+		correlationID = upstreamPeerWatchIDPrefix + uid.String()
+	}
 
 	ctx, cancel := context.WithCancel(ctx)
 	err := s.dataSources.Health.Notify(ctx, &structs.ServiceSpecificRequest{
-		PeerName:   opts.upstreamID.Peer,
+		PeerName:   opts.peer,
 		Datacenter: opts.datacenter,
 		QueryOptions: structs.QueryOptions{
 			Token:  s.token,
@@ -422,6 +453,31 @@ func (s *handlerUpstreams) watchUpstreamTarget(ctx context.Context, snap *Config
 	}
 	snap.WatchedUpstreams[opts.upstreamID][opts.chainID] = cancel
 
+	if uid.Peer == "" {
+		return nil
+	}
+
+	if ok := snap.PeerUpstreamEndpoints.IsWatched(uid); !ok {
+		snap.PeerUpstreamEndpoints.InitWatch(uid, cancel)
+	}
+
+	// Check whether a watch for this peer exists to avoid duplicates.
+	if ok := snap.UpstreamPeerTrustBundles.IsWatched(uid.Peer); !ok {
+		peerCtx, cancel := context.WithCancel(ctx)
+		if err := s.dataSources.TrustBundle.Notify(peerCtx, &cachetype.TrustBundleReadRequest{
+			Request: &pbpeering.TrustBundleReadRequest{
+				Name:      uid.Peer,
+				Partition: uid.PartitionOrDefault(),
+			},
+			QueryOptions: structs.QueryOptions{Token: s.token},
+		}, peerTrustBundleIDPrefix+uid.Peer, s.ch); err != nil {
+			cancel()
+			return fmt.Errorf("error while watching trust bundle for peer %q: %w", uid.Peer, err)
+		}
+
+		snap.UpstreamPeerTrustBundles.InitWatch(uid.Peer, cancel)
+	}
+
 	return nil
 }
 

agent/rpc/peering/service.go

@@ -8,7 +8,6 @@ import (
 	"time"
 
 	"github.com/armon/go-metrics"
-	"github.com/hashicorp/consul/proto/pbpeerstream"
 	"github.com/hashicorp/go-hclog"
 	"github.com/hashicorp/go-memdb"
 	"github.com/hashicorp/go-multierror"
@@ -27,6 +26,7 @@ import (
 	"github.com/hashicorp/consul/agent/structs"
 	"github.com/hashicorp/consul/lib"
 	"github.com/hashicorp/consul/proto/pbpeering"
+	"github.com/hashicorp/consul/proto/pbpeerstream"
 )
 
 var (
@@ -379,6 +379,7 @@ func (s *Server) Establish(
 	}
 
 	var id string
+	serverAddrs := tok.ServerAddresses
 	if existing == nil {
 		id, err = lib.GenerateUUID(s.Backend.CheckPeeringUUID)
 		if err != nil {
@@ -386,6 +387,11 @@ func (s *Server) Establish(
 		}
 	} else {
 		id = existing.ID
+		// If there is a connected stream, assume that the existing ServerAddresses
+		// are up to date and do not try to overwrite them with the token's addresses.
+		if status, ok := s.Tracker.StreamStatus(id); ok && status.Connected {
+			serverAddrs = existing.PeerServerAddresses
+		}
 	}
 
 	// validate that this peer name is not being used as an acceptor already
@@ -397,7 +403,7 @@ func (s *Server) Establish(
 		ID:                  id,
 		Name:                req.PeerName,
 		PeerCAPems:          tok.CA,
-		PeerServerAddresses: tok.ServerAddresses,
+		PeerServerAddresses: serverAddrs,
 		PeerServerName:      tok.ServerName,
 		PeerID:              tok.PeerID,
 		Meta:                req.Meta,
@@ -418,9 +424,9 @@ func (s *Server) Establish(
 	}
 	var exchangeResp *pbpeerstream.ExchangeSecretResponse
 
-	// Loop through the token's addresses once, attempting to fetch the long-lived stream secret.
+	// Loop through the known server addresses once, attempting to fetch the long-lived stream secret.
 	var dialErrors error
-	for _, addr := range peering.PeerServerAddresses {
+	for _, addr := range serverAddrs {
 		exchangeResp, err = exchangeSecret(ctx, addr, tlsOption, &exchangeReq)
 		if err != nil {
 			dialErrors = multierror.Append(dialErrors, fmt.Errorf("failed to exchange peering secret with %q: %w", addr, err))
@@ -720,11 +726,12 @@ func (s *Server) PeeringDelete(ctx context.Context, req *pbpeering.PeeringDelete
 		return nil, err
 	}
 
-	if existing == nil || !existing.IsActive() {
+	if existing == nil || existing.State == pbpeering.PeeringState_DELETING {
 		// Return early when the Peering doesn't exist or is already marked for deletion.
 		// We don't return nil because the pb will fail to marshal.
 		return &pbpeering.PeeringDeleteResponse{}, nil
 	}
+
 	// We are using a write request due to needing to perform a deferred deletion.
 	// The peering gets marked for deletion by setting the DeletedAt field,
 	// and a leader routine will handle deleting the peering.

agent/rpc/peering/service_test.go

@@ -621,38 +621,50 @@ func TestPeeringService_Read_ACLEnforcement(t *testing.T) {
 }
 
 func TestPeeringService_Delete(t *testing.T) {
-	// TODO(peering): see note on newTestServer, refactor to not use this
+	tt := map[string]pbpeering.PeeringState{
-	s := newTestServer(t, nil)
+		"active peering":     pbpeering.PeeringState_ACTIVE,
-
+		"terminated peering": pbpeering.PeeringState_TERMINATED,
-	p := &pbpeering.Peering{
-		ID:                  testUUID(t),
-		Name:                "foo",
-		State:               pbpeering.PeeringState_ESTABLISHING,
-		PeerCAPems:          nil,
-		PeerServerName:      "test",
-		PeerServerAddresses: []string{"addr1"},
 	}
-	err := s.Server.FSM().State().PeeringWrite(10, &pbpeering.PeeringWriteRequest{Peering: p})
-	require.NoError(t, err)
-	require.Nil(t, p.DeletedAt)
-	require.True(t, p.IsActive())
 
-	client := pbpeering.NewPeeringServiceClient(s.ClientConn(t))
+	for name, overrideState := range tt {
+		t.Run(name, func(t *testing.T) {
+			// TODO(peering): see note on newTestServer, refactor to not use this
+			s := newTestServer(t, nil)
 
-	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+			// A pointer is kept for the following peering so that we can modify the object without another PeeringWrite.
-	t.Cleanup(cancel)
+			p := &pbpeering.Peering{
+				ID:                  testUUID(t),
+				Name:                "foo",
+				PeerCAPems:          nil,
+				PeerServerName:      "test",
+				PeerServerAddresses: []string{"addr1"},
+			}
+			err := s.Server.FSM().State().PeeringWrite(10, &pbpeering.PeeringWriteRequest{Peering: p})
+			require.NoError(t, err)
+			require.Nil(t, p.DeletedAt)
+			require.True(t, p.IsActive())
 
-	_, err = client.PeeringDelete(ctx, &pbpeering.PeeringDeleteRequest{Name: "foo"})
+			// Overwrite the peering state to simulate deleting from a non-initial state.
-	require.NoError(t, err)
+			p.State = overrideState
 
-	retry.Run(t, func(r *retry.R) {
+			client := pbpeering.NewPeeringServiceClient(s.ClientConn(t))
-		_, resp, err := s.Server.FSM().State().PeeringRead(nil, state.Query{Value: "foo"})
-		require.NoError(r, err)
 
-		// Initially the peering will be marked for deletion but eventually the leader
+			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
-		// routine will clean it up.
+			t.Cleanup(cancel)
-		require.Nil(r, resp)
+
-	})
+			_, err = client.PeeringDelete(ctx, &pbpeering.PeeringDeleteRequest{Name: "foo"})
+			require.NoError(t, err)
+
+			retry.Run(t, func(r *retry.R) {
+				_, resp, err := s.Server.FSM().State().PeeringRead(nil, state.Query{Value: "foo"})
+				require.NoError(r, err)
+
+				// Initially the peering will be marked for deletion but eventually the leader
+				// routine will clean it up.
+				require.Nil(r, resp)
+			})
+		})
+	}
 }
 
 func TestPeeringService_Delete_ACLEnforcement(t *testing.T) {

agent/sidecar_service.go

@@ -127,9 +127,20 @@ func (a *Agent) sidecarServiceFromNodeService(ns *structs.NodeService, token str
 	if err != nil {
 		return nil, nil, "", err
 	}
-	// Setup default check if none given
+	// Setup default check if none given.
 	if len(checks) < 1 {
-		checks = sidecarDefaultChecks(ns.ID, sidecar.Proxy.LocalServiceAddress, sidecar.Port)
+		// The check should use the sidecar's address because it makes a request to the sidecar.
+		// If the sidecar's address is empty, we fall back to the address of the local service, as set in
+		// sidecar.Proxy.LocalServiceAddress, in the hope that the proxy is also accessible on that address
+		// (which in most cases it is because it's running as a sidecar in the same network).
+		// We could instead fall back to the address of the service as set by (ns.Address), but I've kept it using
+		// sidecar.Proxy.LocalServiceAddress so as to not change things too much in the
+		// process of fixing #14433.
+		checkAddress := sidecar.Address
+		if checkAddress == "" {
+			checkAddress = sidecar.Proxy.LocalServiceAddress
+		}
+		checks = sidecarDefaultChecks(ns.ID, checkAddress, sidecar.Port)
 	}
 
 	return sidecar, checks, token, nil
@@ -202,14 +213,11 @@ func (a *Agent) sidecarPortFromServiceID(sidecarCompoundServiceID structs.Servic
 	return sidecarPort, nil
 }
 
-func sidecarDefaultChecks(serviceID string, localServiceAddress string, port int) []*structs.CheckType {
+func sidecarDefaultChecks(serviceID string, address string, port int) []*structs.CheckType {
-	// Setup default check if none given
 	return []*structs.CheckType{
 		{
-			Name: "Connect Sidecar Listening",
+			Name:     "Connect Sidecar Listening",
-			// Default to localhost rather than agent/service public IP. The checks
+			TCP:      ipaddr.FormatAddressPort(address, port),
-			// can always be overridden if a non-loopback IP is needed.
-			TCP:      ipaddr.FormatAddressPort(localServiceAddress, port),
 			Interval: 10 * time.Second,
 		},
 		{

agent/sidecar_service_test.go

@@ -215,6 +215,141 @@ func TestAgent_sidecarServiceFromNodeService(t *testing.T) {
 			token:   "foo",
 			wantErr: "reserved for internal use",
 		},
+		{
+			name: "uses proxy address for check",
+			sd: &structs.ServiceDefinition{
+				ID:   "web1",
+				Name: "web",
+				Port: 1111,
+				Connect: &structs.ServiceConnect{
+					SidecarService: &structs.ServiceDefinition{
+						Address: "123.123.123.123",
+						Proxy: &structs.ConnectProxyConfig{
+							LocalServiceAddress: "255.255.255.255",
+						},
+					},
+				},
+				Address: "255.255.255.255",
+			},
+			token: "foo",
+			wantNS: &structs.NodeService{
+				EnterpriseMeta:             *structs.DefaultEnterpriseMetaInDefaultPartition(),
+				Kind:                       structs.ServiceKindConnectProxy,
+				ID:                         "web1-sidecar-proxy",
+				Service:                    "web-sidecar-proxy",
+				Port:                       2222,
+				Address:                    "123.123.123.123",
+				LocallyRegisteredAsSidecar: true,
+				Proxy: structs.ConnectProxyConfig{
+					DestinationServiceName: "web",
+					DestinationServiceID:   "web1",
+					LocalServiceAddress:    "255.255.255.255",
+					LocalServicePort:       1111,
+				},
+			},
+			wantChecks: []*structs.CheckType{
+				{
+					Name:     "Connect Sidecar Listening",
+					TCP:      "123.123.123.123:2222",
+					Interval: 10 * time.Second,
+				},
+				{
+					Name:         "Connect Sidecar Aliasing web1",
+					AliasService: "web1",
+				},
+			},
+			wantToken: "foo",
+		},
+		{
+			name: "uses proxy.local_service_address for check if proxy address is empty",
+			sd: &structs.ServiceDefinition{
+				ID:   "web1",
+				Name: "web",
+				Port: 1111,
+				Connect: &structs.ServiceConnect{
+					SidecarService: &structs.ServiceDefinition{
+						Address: "", // Proxy address empty.
+						Proxy: &structs.ConnectProxyConfig{
+							LocalServiceAddress: "1.2.3.4",
+						},
+					},
+				},
+				Address: "", // Service address empty.
+			},
+			token: "foo",
+			wantNS: &structs.NodeService{
+				EnterpriseMeta:             *structs.DefaultEnterpriseMetaInDefaultPartition(),
+				Kind:                       structs.ServiceKindConnectProxy,
+				ID:                         "web1-sidecar-proxy",
+				Service:                    "web-sidecar-proxy",
+				Port:                       2222,
+				Address:                    "",
+				LocallyRegisteredAsSidecar: true,
+				Proxy: structs.ConnectProxyConfig{
+					DestinationServiceName: "web",
+					DestinationServiceID:   "web1",
+					LocalServiceAddress:    "1.2.3.4",
+					LocalServicePort:       1111,
+				},
+			},
+			wantChecks: []*structs.CheckType{
+				{
+					Name:     "Connect Sidecar Listening",
+					TCP:      "1.2.3.4:2222",
+					Interval: 10 * time.Second,
+				},
+				{
+					Name:         "Connect Sidecar Aliasing web1",
+					AliasService: "web1",
+				},
+			},
+			wantToken: "foo",
+		},
+		{
+			name: "uses 127.0.0.1 for check if proxy and proxy.local_service_address are empty",
+			sd: &structs.ServiceDefinition{
+				ID:   "web1",
+				Name: "web",
+				Port: 1111,
+				Connect: &structs.ServiceConnect{
+					SidecarService: &structs.ServiceDefinition{
+						Address: "",
+						Proxy: &structs.ConnectProxyConfig{
+							LocalServiceAddress: "",
+						},
+					},
+				},
+				Address: "",
+			},
+			token: "foo",
+			wantNS: &structs.NodeService{
+				EnterpriseMeta:             *structs.DefaultEnterpriseMetaInDefaultPartition(),
+				Kind:                       structs.ServiceKindConnectProxy,
+				ID:                         "web1-sidecar-proxy",
+				Service:                    "web-sidecar-proxy",
+				Port:                       2222,
+				Address:                    "",
+				LocallyRegisteredAsSidecar: true,
+				Proxy: structs.ConnectProxyConfig{
+					DestinationServiceName: "web",
+					DestinationServiceID:   "web1",
+					LocalServiceAddress:    "127.0.0.1",
+					LocalServicePort:       1111,
+				},
+			},
+			wantChecks: []*structs.CheckType{
+				{
+					Name:     "Connect Sidecar Listening",
+					TCP:      "127.0.0.1:2222",
+					Interval: 10 * time.Second,
+				},
+				{
+					Name:         "Connect Sidecar Aliasing web1",
+					AliasService: "web1",
+				},
+			},
+			wantToken: "foo",
+		},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {

agent/structs/config_entry.go

@@ -3,12 +3,13 @@ package structs
 import (
 	"errors"
 	"fmt"
-	"github.com/miekg/dns"
 	"net"
 	"strconv"
 	"strings"
 	"time"
 
+	"github.com/miekg/dns"
+
 	"github.com/hashicorp/go-multierror"
 	"github.com/mitchellh/hashstructure"
 	"github.com/mitchellh/mapstructure"
@@ -362,6 +363,13 @@ func (e *ProxyConfigEntry) Normalize() error {
 	}
 
 	e.Kind = ProxyDefaults
+
+	// proxy default config only accepts global configs
+	// this check is replicated in normalize() and validate(),
+	// since validate is not called by all the endpoints (e.g., delete)
+	if e.Name != "" && e.Name != ProxyConfigGlobal {
+		return fmt.Errorf("invalid name (%q), only %q is supported", e.Name, ProxyConfigGlobal)
+	}
 	e.Name = ProxyConfigGlobal
 
 	e.EnterpriseMeta.Normalize()
@@ -961,6 +969,11 @@ type PassiveHealthCheck struct {
 	// MaxFailures is the count of consecutive failures that results in a host
 	// being removed from the pool.
 	MaxFailures uint32 `json:",omitempty" alias:"max_failures"`
+
+	// EnforcingConsecutive5xx is the % chance that a host will be actually ejected
+	// when an outlier status is detected through consecutive 5xx.
+	// This setting can be used to disable ejection or to ramp it up slowly. Defaults to 100.
+	EnforcingConsecutive5xx *uint32 `json:",omitempty" alias:"enforcing_consecutive_5xx"`
 }
 
 func (chk *PassiveHealthCheck) Clone() *PassiveHealthCheck {

agent/structs/config_entry_discoverychain.go

@@ -964,11 +964,18 @@ func (e *ServiceResolverConfigEntry) Validate() error {
 
 		// TODO(rb): prevent subsets and default subsets from being defined?
 
-		if r.Service == "" && r.ServiceSubset == "" && r.Namespace == "" && r.Partition == "" && r.Datacenter == "" {
+		if r.isEmpty() {
 			return fmt.Errorf("Redirect is empty")
 		}
 
-		if r.Service == "" {
+		switch {
+		case r.Peer != "" && r.ServiceSubset != "":
+			return fmt.Errorf("Redirect.Peer cannot be set with Redirect.ServiceSubset")
+		case r.Peer != "" && r.Partition != "":
+			return fmt.Errorf("Redirect.Partition cannot be set with Redirect.Peer")
+		case r.Peer != "" && r.Datacenter != "":
+			return fmt.Errorf("Redirect.Peer cannot be set with Redirect.Datacenter")
+		case r.Service == "":
 			if r.ServiceSubset != "" {
 				return fmt.Errorf("Redirect.ServiceSubset defined without Redirect.Service")
 			}
@@ -978,9 +985,12 @@ func (e *ServiceResolverConfigEntry) Validate() error {
 			if r.Partition != "" {
 				return fmt.Errorf("Redirect.Partition defined without Redirect.Service")
 			}
-		} else if r.Service == e.Name {
+			if r.Peer != "" {
-			if r.ServiceSubset != "" && !isSubset(r.ServiceSubset) {
+				return fmt.Errorf("Redirect.Peer defined without Redirect.Service")
-				return fmt.Errorf("Redirect.ServiceSubset %q is not a valid subset of %q", r.ServiceSubset, r.Service)
+			}
+		case r.ServiceSubset != "" && (r.Service == "" || r.Service == e.Name):
+			if !isSubset(r.ServiceSubset) {
+				return fmt.Errorf("Redirect.ServiceSubset %q is not a valid subset of %q", r.ServiceSubset, e.Name)
 			}
 		}
 	}
@@ -1231,6 +1241,25 @@ type ServiceResolverRedirect struct {
 	// Datacenter is the datacenter to resolve the service from instead of the
 	// current one (optional).
 	Datacenter string `json:",omitempty"`
+
+	// Peer is the name of the cluster peer to resolve the service from instead
+	// of the current one (optional).
+	Peer string `json:",omitempty"`
+}
+
+func (r *ServiceResolverRedirect) ToDiscoveryTargetOpts() DiscoveryTargetOpts {
+	return DiscoveryTargetOpts{
+		Service:       r.Service,
+		ServiceSubset: r.ServiceSubset,
+		Namespace:     r.Namespace,
+		Partition:     r.Partition,
+		Datacenter:    r.Datacenter,
+		Peer:          r.Peer,
+	}
+}
+
+func (r *ServiceResolverRedirect) isEmpty() bool {
+	return r.Service == "" && r.ServiceSubset == "" && r.Namespace == "" && r.Partition == "" && r.Datacenter == "" && r.Peer == ""
 }
 
 // There are some restrictions on what is allowed in here:
@@ -1275,6 +1304,14 @@ type ServiceResolverFailover struct {
 	Targets []ServiceResolverFailoverTarget `json:",omitempty"`
 }
 
+func (t *ServiceResolverFailover) ToDiscoveryTargetOpts() DiscoveryTargetOpts {
+	return DiscoveryTargetOpts{
+		Service:       t.Service,
+		ServiceSubset: t.ServiceSubset,
+		Namespace:     t.Namespace,
+	}
+}
+
 func (f *ServiceResolverFailover) isEmpty() bool {
 	return f.Service == "" && f.ServiceSubset == "" && f.Namespace == "" && len(f.Datacenters) == 0 && len(f.Targets) == 0
 }
@@ -1299,6 +1336,17 @@ type ServiceResolverFailoverTarget struct {
 	Peer string `json:",omitempty"`
 }
 
+func (t *ServiceResolverFailoverTarget) ToDiscoveryTargetOpts() DiscoveryTargetOpts {
+	return DiscoveryTargetOpts{
+		Service:       t.Service,
+		ServiceSubset: t.ServiceSubset,
+		Namespace:     t.Namespace,
+		Partition:     t.Partition,
+		Datacenter:    t.Datacenter,
+		Peer:          t.Peer,
+	}
+}
+
 // LoadBalancer determines the load balancing policy and configuration for services
 // issuing requests to this upstream service.
 type LoadBalancer struct {

agent/structs/config_entry_discoverychain_oss_test.go

@@ -72,6 +72,28 @@ func TestServiceResolverConfigEntry_OSS(t *testing.T) {
 			},
 			validateErr: `Bad Failover["*"]: Setting Namespace requires Consul Enterprise`,
 		},
+		{
+			name: "setting redirect Namespace on OSS",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Namespace: "ns1",
+				},
+			},
+			validateErr: `Redirect: Setting Namespace requires Consul Enterprise`,
+		},
+		{
+			name: "setting redirect Partition on OSS",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Partition: "ap1",
+				},
+			},
+			validateErr: `Redirect: Setting Partition requires Consul Enterprise`,
+		},
 	}
 
 	// Bulk add a bunch of similar validation cases.

agent/structs/config_entry_discoverychain_test.go

@@ -655,6 +655,41 @@ func TestServiceResolverConfigEntry(t *testing.T) {
 			},
 			validateErr: `Redirect.ServiceSubset "gone" is not a valid subset of "test"`,
 		},
+		{
+			name: "redirect with peer and subset",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Peer:          "cluster-01",
+					ServiceSubset: "gone",
+				},
+			},
+			validateErr: `Redirect.Peer cannot be set with Redirect.ServiceSubset`,
+		},
+		{
+			name: "redirect with peer and datacenter",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Peer:       "cluster-01",
+					Datacenter: "dc2",
+				},
+			},
+			validateErr: `Redirect.Peer cannot be set with Redirect.Datacenter`,
+		},
+		{
+			name: "redirect with peer and datacenter",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Peer: "cluster-01",
+				},
+			},
+			validateErr: `Redirect.Peer defined without Redirect.Service`,
+		},
 		{
 			name: "self redirect with valid subset",
 			entry: &ServiceResolverConfigEntry{
@@ -669,6 +704,17 @@ func TestServiceResolverConfigEntry(t *testing.T) {
 				},
 			},
 		},
+		{
+			name: "redirect to peer",
+			entry: &ServiceResolverConfigEntry{
+				Kind: ServiceResolver,
+				Name: "test",
+				Redirect: &ServiceResolverRedirect{
+					Service: "other",
+					Peer:    "cluster-01",
+				},
+			},
+		},
 		{
 			name: "simple wildcard failover",
 			entry: &ServiceResolverConfigEntry{

agent/structs/config_entry_test.go

@@ -2754,8 +2754,9 @@ func TestUpstreamConfig_MergeInto(t *testing.T) {
 					MaxConcurrentRequests: intPointer(12),
 				},
 				"passive_health_check": &PassiveHealthCheck{
-					MaxFailures: 13,
+					MaxFailures:             13,
-					Interval:    14 * time.Second,
+					Interval:                14 * time.Second,
+					EnforcingConsecutive5xx: uintPointer(80),
 				},
 				"mesh_gateway": MeshGatewayConfig{Mode: MeshGatewayModeLocal},
 			},
@@ -2770,8 +2771,9 @@ func TestUpstreamConfig_MergeInto(t *testing.T) {
 					MaxConcurrentRequests: intPointer(12),
 				},
 				"passive_health_check": &PassiveHealthCheck{
-					MaxFailures: 13,
+					MaxFailures:             13,
-					Interval:    14 * time.Second,
+					Interval:                14 * time.Second,
+					EnforcingConsecutive5xx: uintPointer(80),
 				},
 				"mesh_gateway": MeshGatewayConfig{Mode: MeshGatewayModeLocal},
 			},
@@ -2944,6 +2946,28 @@ func TestParseUpstreamConfig(t *testing.T) {
 	}
 }
 
+func TestProxyConfigEntry(t *testing.T) {
+	cases := map[string]configEntryTestcase{
+		"proxy config name provided is not global": {
+			entry: &ProxyConfigEntry{
+				Name: "foo",
+			},
+			normalizeErr: `invalid name ("foo"), only "global" is supported`,
+		},
+		"proxy config has no name": {
+			entry: &ProxyConfigEntry{
+				Name: "",
+			},
+			expected: &ProxyConfigEntry{
+				Name:           ProxyConfigGlobal,
+				Kind:           ProxyDefaults,
+				EnterpriseMeta: *acl.DefaultEnterpriseMeta(),
+			},
+		},
+	}
+	testConfigEntryNormalizeAndValidate(t, cases)
+}
+
 func requireContainsLower(t *testing.T, haystack, needle string) {
 	t.Helper()
 	require.Contains(t, strings.ToLower(haystack), strings.ToLower(needle))
@@ -3046,3 +3070,7 @@ func testConfigEntryNormalizeAndValidate(t *testing.T, cases map[string]configEn
 		})
 	}
 }
+
+func uintPointer(v uint32) *uint32 {
+	return &v
+}

agent/structs/discovery_chain.go

@@ -56,7 +56,12 @@ type CompiledDiscoveryChain struct {
 // ID returns an ID that encodes the service, namespace, partition, and datacenter.
 // This ID allows us to compare a discovery chain target to the chain upstream itself.
 func (c *CompiledDiscoveryChain) ID() string {
-	return chainID("", c.ServiceName, c.Namespace, c.Partition, c.Datacenter)
+	return chainID(DiscoveryTargetOpts{
+		Service:    c.ServiceName,
+		Namespace:  c.Namespace,
+		Partition:  c.Partition,
+		Datacenter: c.Datacenter,
+	})
 }
 
 func (c *CompiledDiscoveryChain) CompoundServiceName() ServiceName {
@@ -185,6 +190,7 @@ type DiscoveryTarget struct {
 	Namespace     string `json:",omitempty"`
 	Partition     string `json:",omitempty"`
 	Datacenter    string `json:",omitempty"`
+	Peer          string `json:",omitempty"`
 
 	MeshGateway MeshGatewayConfig     `json:",omitempty"`
 	Subset      ServiceResolverSubset `json:",omitempty"`
@@ -240,28 +246,52 @@ func (t *DiscoveryTarget) UnmarshalJSON(data []byte) error {
 	return nil
 }
 
-func NewDiscoveryTarget(service, serviceSubset, namespace, partition, datacenter string) *DiscoveryTarget {
+type DiscoveryTargetOpts struct {
+	Service       string
+	ServiceSubset string
+	Namespace     string
+	Partition     string
+	Datacenter    string
+	Peer          string
+}
+
+func NewDiscoveryTarget(opts DiscoveryTargetOpts) *DiscoveryTarget {
 	t := &DiscoveryTarget{
-		Service:       service,
+		Service:       opts.Service,
-		ServiceSubset: serviceSubset,
+		ServiceSubset: opts.ServiceSubset,
-		Namespace:     namespace,
+		Namespace:     opts.Namespace,
-		Partition:     partition,
+		Partition:     opts.Partition,
-		Datacenter:    datacenter,
+		Datacenter:    opts.Datacenter,
+		Peer:          opts.Peer,
 	}
 	t.setID()
 	return t
 }
 
-func chainID(subset, service, namespace, partition, dc string) string {
+func (t *DiscoveryTarget) ToDiscoveryTargetOpts() DiscoveryTargetOpts {
-	// NOTE: this format is similar to the SNI syntax for simplicity
+	return DiscoveryTargetOpts{
-	if subset == "" {
+		Service:       t.Service,
-		return fmt.Sprintf("%s.%s.%s.%s", service, namespace, partition, dc)
+		ServiceSubset: t.ServiceSubset,
+		Namespace:     t.Namespace,
+		Partition:     t.Partition,
+		Datacenter:    t.Datacenter,
+		Peer:          t.Peer,
 	}
-	return fmt.Sprintf("%s.%s.%s.%s.%s", subset, service, namespace, partition, dc)
+}
+
+func chainID(opts DiscoveryTargetOpts) string {
+	// NOTE: this format is similar to the SNI syntax for simplicity
+	if opts.Peer != "" {
+		return fmt.Sprintf("%s.%s.default.external.%s", opts.Service, opts.Namespace, opts.Peer)
+	}
+	if opts.ServiceSubset == "" {
+		return fmt.Sprintf("%s.%s.%s.%s", opts.Service, opts.Namespace, opts.Partition, opts.Datacenter)
+	}
+	return fmt.Sprintf("%s.%s.%s.%s.%s", opts.ServiceSubset, opts.Service, opts.Namespace, opts.Partition, opts.Datacenter)
 }
 
 func (t *DiscoveryTarget) setID() {
-	t.ID = chainID(t.ServiceSubset, t.Service, t.Namespace, t.Partition, t.Datacenter)
+	t.ID = chainID(t.ToDiscoveryTargetOpts())
 }
 
 func (t *DiscoveryTarget) String() string {

agent/structs/testing_catalog.go

@@ -53,6 +53,28 @@ func TestNodeServiceWithName(t testing.T, name string) *NodeService {
 	}
 }
 
+const peerTrustDomain = "1c053652-8512-4373-90cf-5a7f6263a994.consul"
+
+func TestNodeServiceWithNameInPeer(t testing.T, name string, peer string) *NodeService {
+	service := "payments"
+	return &NodeService{
+		Kind:    ServiceKindTypical,
+		Service: name,
+		Port:    8080,
+		Connect: ServiceConnect{
+			PeerMeta: &PeeringServiceMeta{
+				SNI: []string{
+					service + ".default.default." + peer + ".external." + peerTrustDomain,
+				},
+				SpiffeID: []string{
+					"spiffe://" + peerTrustDomain + "/ns/default/dc/" + peer + "-dc/svc/" + service,
+				},
+				Protocol: "tcp",
+			},
+		},
+	}
+}
+
 // TestNodeServiceProxy returns a *NodeService representing a valid
 // Connect proxy.
 func TestNodeServiceProxy(t testing.T) *NodeService {

agent/structs/testing_connect_proxy_config.go

@@ -26,7 +26,7 @@ func TestUpstreams(t testing.T) Upstreams {
 			Config: map[string]interface{}{
 				// Float because this is how it is decoded by JSON decoder so this
 				// enables the value returned to be compared directly to a decoded JSON
-				// response without spurios type loss.
+				// response without spurious type loss.
 				"connect_timeout_ms": float64(1000),
 			},
 		},

agent/submatview/local_materializer.go

@@ -66,6 +66,10 @@ func (m *LocalMaterializer) Run(ctx context.Context) {
 		if ctx.Err() != nil {
 			return
 		}
+		if m.isTerminalError(err) {
+			return
+		}
+
 		m.mat.handleError(req, err)
 
 		if err := m.mat.retryWaiter.Wait(ctx); err != nil {
@@ -74,6 +78,14 @@ func (m *LocalMaterializer) Run(ctx context.Context) {
 	}
 }
 
+// isTerminalError determines whether the given error cannot be recovered from
+// and should cause the materializer to halt and be evicted from the view store.
+//
+// This roughly matches the logic in agent/proxycfg-glue.newUpdateEvent.
+func (m *LocalMaterializer) isTerminalError(err error) bool {
+	return acl.IsErrNotFound(err)
+}
+
 // subscribeOnce opens a new subscription to a local backend and runs
 // for its lifetime or until the view is closed.
 func (m *LocalMaterializer) subscribeOnce(ctx context.Context, req *pbsubscribe.SubscribeRequest) error {

agent/submatview/store.go

@@ -47,6 +47,9 @@ type entry struct {
 	// requests is the count of active requests using this entry. This entry will
 	// remain in the store as long as this count remains > 0.
 	requests int
+	// evicting is used to mark an entry that will be evicted when the current in-
+	// flight requests finish.
+	evicting bool
 }
 
 // NewStore creates and returns a Store that is ready for use. The caller must
@@ -89,6 +92,7 @@ func (s *Store) Run(ctx context.Context) {
 
 			// Only stop the materializer if there are no active requests.
 			if e.requests == 0 {
+				s.logger.Trace("evicting item from store", "key", he.Key())
 				e.stop()
 				delete(s.byKey, he.Key())
 			}
@@ -187,13 +191,13 @@ func (s *Store) NotifyCallback(
 					"error", err,
 					"request-type", req.Type(),
 					"index", index)
-				continue
 			}
 
 			index = result.Index
 			cb(ctx, cache.UpdateEvent{
 				CorrelationID: correlationID,
 				Result:        result.Value,
+				Err:           err,
 				Meta:          cache.ResultMeta{Index: result.Index, Hit: result.Cached},
 			})
 		}
@@ -211,6 +215,9 @@ func (s *Store) readEntry(req Request) (string, Materializer, error) {
 	defer s.lock.Unlock()
 	e, ok := s.byKey[key]
 	if ok {
+		if e.evicting {
+			return "", nil, errors.New("item is marked for eviction")
+		}
 		e.requests++
 		s.byKey[key] = e
 		return key, e.materializer, nil
@@ -222,7 +229,18 @@ func (s *Store) readEntry(req Request) (string, Materializer, error) {
 	}
 
 	ctx, cancel := context.WithCancel(context.Background())
-	go mat.Run(ctx)
+	go func() {
+		mat.Run(ctx)
+
+		// Materializers run until they either reach their TTL and are evicted (which
+		// cancels the given context) or encounter an irrecoverable error.
+		//
+		// If the context hasn't been canceled, we know it's the error case so we
+		// trigger an immediate eviction.
+		if ctx.Err() == nil {
+			s.evictNow(key)
+		}
+	}()
 
 	e = entry{
 		materializer: mat,
@@ -233,6 +251,28 @@ func (s *Store) readEntry(req Request) (string, Materializer, error) {
 	return key, e.materializer, nil
 }
 
+// evictNow causes the item with the given key to be evicted immediately.
+//
+// If there are requests in-flight, the item is marked for eviction such that
+// once the requests have been served releaseEntry will move it to the top of
+// the expiry heap. If there are no requests in-flight, evictNow will move the
+// item to the top of the expiry heap itself.
+//
+// In either case, the entry's evicting flag prevents it from being served by
+// readEntry (and thereby gaining new in-flight requests).
+func (s *Store) evictNow(key string) {
+	s.lock.Lock()
+	defer s.lock.Unlock()
+
+	e := s.byKey[key]
+	e.evicting = true
+	s.byKey[key] = e
+
+	if e.requests == 0 {
+		s.expireNowLocked(key)
+	}
+}
+
 // releaseEntry decrements the request count and starts an expiry timer if the
 // count has reached 0. Must be called once for every call to readEntry.
 func (s *Store) releaseEntry(key string) {
@@ -246,6 +286,11 @@ func (s *Store) releaseEntry(key string) {
 		return
 	}
 
+	if e.evicting {
+		s.expireNowLocked(key)
+		return
+	}
+
 	if e.expiry.Index() == ttlcache.NotIndexed {
 		e.expiry = s.expiryHeap.Add(key, s.idleTTL)
 		s.byKey[key] = e
@@ -255,6 +300,17 @@ func (s *Store) releaseEntry(key string) {
 	s.expiryHeap.Update(e.expiry.Index(), s.idleTTL)
 }
 
+// expireNowLocked moves the item with the given key to the top of the expiry
+// heap, causing it to be picked up by the expiry loop and evicted immediately.
+func (s *Store) expireNowLocked(key string) {
+	e := s.byKey[key]
+	if idx := e.expiry.Index(); idx != ttlcache.NotIndexed {
+		s.expiryHeap.Remove(idx)
+	}
+	e.expiry = s.expiryHeap.Add(key, time.Duration(0))
+	s.byKey[key] = e
+}
+
 // makeEntryKey matches agent/cache.makeEntryKey, but may change in the future.
 func makeEntryKey(typ string, r cache.RequestInfo) string {
 	return fmt.Sprintf("%s/%s/%s/%s", typ, r.Datacenter, r.Token, r.Key)

agent/submatview/store_test.go

@@ -509,3 +509,75 @@ func TestStore_Run_ExpiresEntries(t *testing.T) {
 	require.Len(t, store.byKey, 0)
 	require.Equal(t, ttlcache.NotIndexed, e.expiry.Index())
 }
+
+func TestStore_Run_FailingMaterializer(t *testing.T) {
+	ctx, cancel := context.WithCancel(context.Background())
+	t.Cleanup(cancel)
+
+	store := NewStore(hclog.NewNullLogger())
+	store.idleTTL = 24 * time.Hour
+	go store.Run(ctx)
+
+	t.Run("with an in-flight request", func(t *testing.T) {
+		req := &failingMaterializerRequest{
+			doneCh: make(chan struct{}),
+		}
+
+		ch := make(chan cache.UpdateEvent)
+		reqCtx, reqCancel := context.WithCancel(context.Background())
+		t.Cleanup(reqCancel)
+		require.NoError(t, store.Notify(reqCtx, req, "", ch))
+
+		assertRequestCount(t, store, req, 1)
+
+		// Cause the materializer to "fail" (exit before its context is canceled).
+		close(req.doneCh)
+
+		// End the in-flight request.
+		reqCancel()
+
+		// Check that the item was evicted.
+		retry.Run(t, func(r *retry.R) {
+			store.lock.Lock()
+			defer store.lock.Unlock()
+
+			require.Len(r, store.byKey, 0)
+		})
+	})
+
+	t.Run("with no in-flight requests", func(t *testing.T) {
+		req := &failingMaterializerRequest{
+			doneCh: make(chan struct{}),
+		}
+
+		// Cause the materializer to "fail" (exit before its context is canceled).
+		close(req.doneCh)
+
+		// Check that the item was evicted.
+		retry.Run(t, func(r *retry.R) {
+			store.lock.Lock()
+			defer store.lock.Unlock()
+
+			require.Len(r, store.byKey, 0)
+		})
+	})
+}
+
+type failingMaterializerRequest struct {
+	doneCh chan struct{}
+}
+
+func (failingMaterializerRequest) CacheInfo() cache.RequestInfo { return cache.RequestInfo{} }
+func (failingMaterializerRequest) Type() string                 { return "test.FailingMaterializerRequest" }
+
+func (r *failingMaterializerRequest) NewMaterializer() (Materializer, error) {
+	return &failingMaterializer{doneCh: r.doneCh}, nil
+}
+
+type failingMaterializer struct {
+	doneCh <-chan struct{}
+}
+
+func (failingMaterializer) Query(context.Context, uint64) (Result, error) { return Result{}, nil }
+
+func (m *failingMaterializer) Run(context.Context) { <-m.doneCh }

agent/txn_endpoint.go

@@ -185,6 +185,7 @@ func (s *HTTPHandlers) convertOps(resp http.ResponseWriter, req *http.Request) (
 						Address:         node.Address,
 						Datacenter:      node.Datacenter,
 						TaggedAddresses: node.TaggedAddresses,
+						PeerName:        node.PeerName,
 						Meta:            node.Meta,
 						RaftIndex: structs.RaftIndex{
 							ModifyIndex: node.ModifyIndex,
@@ -207,6 +208,7 @@ func (s *HTTPHandlers) convertOps(resp http.ResponseWriter, req *http.Request) (
 					Service: structs.NodeService{
 						ID:      svc.ID,
 						Service: svc.Service,
+						Kind:    structs.ServiceKind(svc.Kind),
 						Tags:    svc.Tags,
 						Address: svc.Address,
 						Meta:    svc.Meta,
@@ -226,6 +228,39 @@ func (s *HTTPHandlers) convertOps(resp http.ResponseWriter, req *http.Request) (
 					},
 				},
 			}
+
+			if svc.Proxy != nil {
+				out.Service.Service.Proxy = structs.ConnectProxyConfig{}
+				t := &out.Service.Service.Proxy
+				if svc.Proxy.DestinationServiceName != "" {
+					t.DestinationServiceName = svc.Proxy.DestinationServiceName
+				}
+				if svc.Proxy.DestinationServiceID != "" {
+					t.DestinationServiceID = svc.Proxy.DestinationServiceID
+				}
+				if svc.Proxy.LocalServiceAddress != "" {
+					t.LocalServiceAddress = svc.Proxy.LocalServiceAddress
+				}
+				if svc.Proxy.LocalServicePort != 0 {
+					t.LocalServicePort = svc.Proxy.LocalServicePort
+				}
+				if svc.Proxy.LocalServiceSocketPath != "" {
+					t.LocalServiceSocketPath = svc.Proxy.LocalServiceSocketPath
+				}
+				if svc.Proxy.MeshGateway.Mode != "" {
+					t.MeshGateway.Mode = structs.MeshGatewayMode(svc.Proxy.MeshGateway.Mode)
+				}
+
+				if svc.Proxy.TransparentProxy != nil {
+					if svc.Proxy.TransparentProxy.DialedDirectly {
+						t.TransparentProxy.DialedDirectly = svc.Proxy.TransparentProxy.DialedDirectly
+					}
+
+					if svc.Proxy.TransparentProxy.OutboundListenerPort != 0 {
+						t.TransparentProxy.OutboundListenerPort = svc.Proxy.TransparentProxy.OutboundListenerPort
+					}
+				}
+			}
 			opsRPC = append(opsRPC, out)
 
 		case in.Check != nil:
@@ -265,6 +300,8 @@ func (s *HTTPHandlers) convertOps(resp http.ResponseWriter, req *http.Request) (
 						ServiceID:   check.ServiceID,
 						ServiceName: check.ServiceName,
 						ServiceTags: check.ServiceTags,
+						PeerName:    check.PeerName,
+						ExposedPort: check.ExposedPort,
 						Definition: structs.HealthCheckDefinition{
 							HTTP:                           check.Definition.HTTP,
 							TLSServerName:                  check.Definition.TLSServerName,

agent/txn_endpoint_test.go

@@ -585,6 +585,7 @@ func TestTxnEndpoint_UpdateCheck(t *testing.T) {
 				"Output": "success",
 				"ServiceID": "",
 				"ServiceName": "",
+				"ExposedPort": 5678,
 				"Definition": {
 					"IntervalDuration": "15s",
 					"TimeoutDuration": "15s",
@@ -600,12 +601,8 @@ func TestTxnEndpoint_UpdateCheck(t *testing.T) {
 	req, _ := http.NewRequest("PUT", "/v1/txn", buf)
 	resp := httptest.NewRecorder()
 	obj, err := a.srv.Txn(resp, req)
-	if err != nil {
+	require.NoError(t, err)
-		t.Fatalf("err: %v", err)
+	require.Equal(t, 200, resp.Code, resp.Body)
-	}
-	if resp.Code != 200 {
-		t.Fatalf("expected 200, got %d", resp.Code)
-	}
 
 	txnResp, ok := obj.(structs.TxnResponse)
 	if !ok {
@@ -662,12 +659,13 @@ func TestTxnEndpoint_UpdateCheck(t *testing.T) {
 			},
 			&structs.TxnResult{
 				Check: &structs.HealthCheck{
-					Node:    a.config.NodeName,
+					Node:        a.config.NodeName,
-					CheckID: "nodecheck",
+					CheckID:     "nodecheck",
-					Name:    "Node http check",
+					Name:        "Node http check",
-					Status:  api.HealthPassing,
+					Status:      api.HealthPassing,
-					Notes:   "Http based health check",
+					Notes:       "Http based health check",
-					Output:  "success",
+					Output:      "success",
+					ExposedPort: 5678,
 					Definition: structs.HealthCheckDefinition{
 						Interval:                       15 * time.Second,
 						Timeout:                        15 * time.Second,
@@ -686,3 +684,117 @@ func TestTxnEndpoint_UpdateCheck(t *testing.T) {
 	}
 	assert.Equal(t, expected, txnResp)
 }
+
+func TestTxnEndpoint_NodeService(t *testing.T) {
+	if testing.Short() {
+		t.Skip("too slow for testing.Short")
+	}
+
+	t.Parallel()
+	a := NewTestAgent(t, "")
+	defer a.Shutdown()
+	testrpc.WaitForTestAgent(t, a.RPC, "dc1")
+
+	// Make sure the fields of a check are handled correctly when both creating and
+	// updating, and test both sets of duration fields to ensure backwards compatibility.
+	buf := bytes.NewBuffer([]byte(fmt.Sprintf(`
+[
+	{
+		"Service": {
+		  	"Verb": "set",
+		  	"Node": "%s",
+		  	"Service": {
+				"Service": "test",
+				"Port": 4444
+		  	}
+		}
+	},
+	{
+		"Service": {
+			"Verb": "set",
+			"Node": "%s",
+			"Service": {
+				"Service": "test-sidecar-proxy",
+				"Port": 20000,
+				"Kind": "connect-proxy",
+				"Proxy": {
+				  	"DestinationServiceName": "test",
+				  	"DestinationServiceID": "test",
+				  	"LocalServiceAddress": "127.0.0.1",
+				  	"LocalServicePort": 4444,
+				  	"upstreams": [
+						{
+							"DestinationName": "fake-backend",
+							"LocalBindPort": 25001
+						}
+				  	]
+				}
+			}
+		}
+	}
+]
+`, a.config.NodeName, a.config.NodeName)))
+	req, _ := http.NewRequest("PUT", "/v1/txn", buf)
+	resp := httptest.NewRecorder()
+	obj, err := a.srv.Txn(resp, req)
+	require.NoError(t, err)
+	require.Equal(t, 200, resp.Code)
+
+	txnResp, ok := obj.(structs.TxnResponse)
+	if !ok {
+		t.Fatalf("bad type: %T", obj)
+	}
+	require.Equal(t, 2, len(txnResp.Results))
+
+	index := txnResp.Results[0].Service.ModifyIndex
+	expected := structs.TxnResponse{
+		Results: structs.TxnResults{
+			&structs.TxnResult{
+				Service: &structs.NodeService{
+					Service: "test",
+					ID:      "test",
+					Port:    4444,
+					Weights: &structs.Weights{
+						Passing: 1,
+						Warning: 1,
+					},
+					RaftIndex: structs.RaftIndex{
+						CreateIndex: index,
+						ModifyIndex: index,
+					},
+					EnterpriseMeta: *structs.DefaultEnterpriseMetaInDefaultPartition(),
+				},
+			},
+			&structs.TxnResult{
+				Service: &structs.NodeService{
+					Service: "test-sidecar-proxy",
+					ID:      "test-sidecar-proxy",
+					Port:    20000,
+					Kind:    "connect-proxy",
+					Weights: &structs.Weights{
+						Passing: 1,
+						Warning: 1,
+					},
+					Proxy: structs.ConnectProxyConfig{
+						DestinationServiceName: "test",
+						DestinationServiceID:   "test",
+						LocalServiceAddress:    "127.0.0.1",
+						LocalServicePort:       4444,
+					},
+					TaggedAddresses: map[string]structs.ServiceAddress{
+						"consul-virtual": {
+							Address: "240.0.0.1",
+							Port:    20000,
+						},
+					},
+					RaftIndex: structs.RaftIndex{
+						CreateIndex: index,
+						ModifyIndex: index,
+					},
+					EnterpriseMeta: *structs.DefaultEnterpriseMetaInDefaultPartition(),
+				},
+			},
+		},
+	}
+	assert.Equal(t, expected, txnResp)
+}

agent/ui_endpoint.go

@@ -211,7 +211,9 @@ func (s *HTTPHandlers) UIServices(resp http.ResponseWriter, req *http.Request) (
 	if done := s.parse(resp, req, &args.Datacenter, &args.QueryOptions); done {
 		return nil, nil
 	}
-
+	if peer := req.URL.Query().Get("peer"); peer != "" {
+		args.PeerName = peer
+	}
 	if err := s.parseEntMeta(req, &args.EnterpriseMeta); err != nil {
 		return nil, err
 	}

agent/xds/clusters.go

@@ -88,29 +88,26 @@ func (s *ResourceGenerator) clustersFromSnapshotConnectProxy(cfgSnap *proxycfg.C
 		clusters = append(clusters, passthroughs...)
 	}
 
-	// NOTE: Any time we skip a chain below we MUST also skip that discovery chain in endpoints.go
+	getUpstream := func(uid proxycfg.UpstreamID) (*structs.Upstream, bool) {
-	// so that the sets of endpoints generated matches the sets of clusters.
-	for uid, chain := range cfgSnap.ConnectProxy.DiscoveryChain {
 		upstream := cfgSnap.ConnectProxy.UpstreamConfig[uid]
 
 		explicit := upstream.HasLocalPortOrSocket()
 		implicit := cfgSnap.ConnectProxy.IsImplicitUpstream(uid)
-		if !implicit && !explicit {
+		return upstream, !implicit && !explicit
-			// Discovery chain is not associated with a known explicit or implicit upstream so it is skipped.
+	}
-			continue
-		}
 
-		chainEndpoints, ok := cfgSnap.ConnectProxy.WatchedUpstreamEndpoints[uid]
+	// NOTE: Any time we skip a chain below we MUST also skip that discovery chain in endpoints.go
-		if !ok {
+	// so that the sets of endpoints generated matches the sets of clusters.
-			// this should not happen
+	for uid, chain := range cfgSnap.ConnectProxy.DiscoveryChain {
-			return nil, fmt.Errorf("no endpoint map for upstream %q", uid)
+		upstream, skip := getUpstream(uid)
+		if skip {
+			continue
 		}
 
 		upstreamClusters, err := s.makeUpstreamClustersForDiscoveryChain(
 			uid,
 			upstream,
 			chain,
-			chainEndpoints,
 			cfgSnap,
 			false,
 		)
@@ -127,18 +124,15 @@ func (s *ResourceGenerator) clustersFromSnapshotConnectProxy(cfgSnap *proxycfg.C
 	// upstream in endpoints.go so that the sets of endpoints generated matches
 	// the sets of clusters.
 	for _, uid := range cfgSnap.ConnectProxy.PeeredUpstreamIDs() {
-		upstreamCfg := cfgSnap.ConnectProxy.UpstreamConfig[uid]
+		upstream, skip := getUpstream(uid)
-
+		if skip {
-		explicit := upstreamCfg.HasLocalPortOrSocket()
-		implicit := cfgSnap.ConnectProxy.IsImplicitUpstream(uid)
-		if !implicit && !explicit {
-			// Not associated with a known explicit or implicit upstream so it is skipped.
 			continue
 		}
 
 		peerMeta := cfgSnap.ConnectProxy.UpstreamPeerMeta(uid)
+		cfg := s.getAndModifyUpstreamConfigForPeeredListener(uid, upstream, peerMeta)
 
-		upstreamCluster, err := s.makeUpstreamClusterForPeerService(uid, upstreamCfg, peerMeta, cfgSnap)
+		upstreamCluster, err := s.makeUpstreamClusterForPeerService(uid, cfg, peerMeta, cfgSnap)
 		if err != nil {
 			return nil, err
 		}
@@ -652,17 +646,10 @@ func (s *ResourceGenerator) clustersFromSnapshotIngressGateway(cfgSnap *proxycfg
 				return nil, fmt.Errorf("no discovery chain for upstream %q", uid)
 			}
 
-			chainEndpoints, ok := cfgSnap.IngressGateway.WatchedUpstreamEndpoints[uid]
-			if !ok {
-				// this should not happen
-				return nil, fmt.Errorf("no endpoint map for upstream %q", uid)
-			}
-
 			upstreamClusters, err := s.makeUpstreamClustersForDiscoveryChain(
 				uid,
 				&u,
 				chain,
-				chainEndpoints,
 				cfgSnap,
 				false,
 			)
@@ -745,7 +732,7 @@ func (s *ResourceGenerator) makeAppCluster(cfgSnap *proxycfg.ConfigSnapshot, nam
 
 func (s *ResourceGenerator) makeUpstreamClusterForPeerService(
 	uid proxycfg.UpstreamID,
-	upstream *structs.Upstream,
+	upstreamConfig structs.UpstreamConfig,
 	peerMeta structs.PeeringServiceMeta,
 	cfgSnap *proxycfg.ConfigSnapshot,
 ) (*envoy_cluster_v3.Cluster, error) {
@@ -754,16 +741,21 @@ func (s *ResourceGenerator) makeUpstreamClusterForPeerService(
 		err error
 	)
 
-	cfg := s.getAndModifyUpstreamConfigForPeeredListener(uid, upstream, peerMeta)
+	if upstreamConfig.EnvoyClusterJSON != "" {
-	if cfg.EnvoyClusterJSON != "" {
+		c, err = makeClusterFromUserConfig(upstreamConfig.EnvoyClusterJSON)
-		c, err = makeClusterFromUserConfig(cfg.EnvoyClusterJSON)
 		if err != nil {
 			return c, err
 		}
 		// In the happy path don't return yet as we need to inject TLS config still.
 	}
 
-	tbs, ok := cfgSnap.ConnectProxy.UpstreamPeerTrustBundles.Get(uid.Peer)
+	upstreamsSnapshot, err := cfgSnap.ToConfigSnapshotUpstreams()
+
+	if err != nil {
+		return c, err
+	}
+
+	tbs, ok := upstreamsSnapshot.UpstreamPeerTrustBundles.Get(uid.Peer)
 	if !ok {
 		// this should never happen since we loop through upstreams with
 		// set trust bundles
@@ -772,22 +764,29 @@ func (s *ResourceGenerator) makeUpstreamClusterForPeerService(
 
 	clusterName := generatePeeredClusterName(uid, tbs)
 
+	outlierDetection := ToOutlierDetection(upstreamConfig.PassiveHealthCheck)
+	// We can't rely on health checks for services on cluster peers because they
+	// don't take into account service resolvers, splitters and routers. Setting
+	// MaxEjectionPercent too 100% gives outlier detection the power to eject the
+	// entire cluster.
+	outlierDetection.MaxEjectionPercent = &wrappers.UInt32Value{Value: 100}
+
 	s.Logger.Trace("generating cluster for", "cluster", clusterName)
 	if c == nil {
 		c = &envoy_cluster_v3.Cluster{
 			Name:           clusterName,
-			ConnectTimeout: durationpb.New(time.Duration(cfg.ConnectTimeoutMs) * time.Millisecond),
+			ConnectTimeout: durationpb.New(time.Duration(upstreamConfig.ConnectTimeoutMs) * time.Millisecond),
 			CommonLbConfig: &envoy_cluster_v3.Cluster_CommonLbConfig{
 				HealthyPanicThreshold: &envoy_type_v3.Percent{
 					Value: 0, // disable panic threshold
 				},
 			},
 			CircuitBreakers: &envoy_cluster_v3.CircuitBreakers{
-				Thresholds: makeThresholdsIfNeeded(cfg.Limits),
+				Thresholds: makeThresholdsIfNeeded(upstreamConfig.Limits),
 			},
-			OutlierDetection: ToOutlierDetection(cfg.PassiveHealthCheck),
+			OutlierDetection: outlierDetection,
 		}
-		if cfg.Protocol == "http2" || cfg.Protocol == "grpc" {
+		if upstreamConfig.Protocol == "http2" || upstreamConfig.Protocol == "grpc" {
 			if err := s.setHttp2ProtocolOptions(c); err != nil {
 				return c, err
 			}
@@ -821,12 +820,11 @@ func (s *ResourceGenerator) makeUpstreamClusterForPeerService(
 				false, /*onlyPassing*/
 			)
 		}
-
 	}
 
 	rootPEMs := cfgSnap.RootPEMs()
 	if uid.Peer != "" {
-		tbs, _ := cfgSnap.ConnectProxy.UpstreamPeerTrustBundles.Get(uid.Peer)
+		tbs, _ := upstreamsSnapshot.UpstreamPeerTrustBundles.Get(uid.Peer)
 		rootPEMs = tbs.ConcatenatedRootPEMs()
 	}
 
@@ -961,7 +959,6 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 	uid proxycfg.UpstreamID,
 	upstream *structs.Upstream,
 	chain *structs.CompiledDiscoveryChain,
-	chainEndpoints map[string]structs.CheckServiceNodes,
 	cfgSnap *proxycfg.ConfigSnapshot,
 	forMeshGateway bool,
 ) ([]*envoy_cluster_v3.Cluster, error) {
@@ -978,7 +975,15 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 		upstreamConfigMap = upstream.Config
 	}
 
-	cfg, err := structs.ParseUpstreamConfigNoDefaults(upstreamConfigMap)
+	upstreamsSnapshot, err := cfgSnap.ToConfigSnapshotUpstreams()
+
+	// Mesh gateways are exempt because upstreamsSnapshot is only used for
+	// cluster peering targets and transative failover/redirects are unsupported.
+	if err != nil && !forMeshGateway {
+		return nil, fmt.Errorf("No upstream snapshot for gateway mode %q", cfgSnap.Kind)
+	}
+
+	rawUpstreamConfig, err := structs.ParseUpstreamConfigNoDefaults(upstreamConfigMap)
 	if err != nil {
 		// Don't hard fail on a config typo, just warn. The parse func returns
 		// default config if there is an error so it's safe to continue.
@@ -986,13 +991,28 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 			"error", err)
 	}
 
+	finalizeUpstreamConfig := func(cfg structs.UpstreamConfig, connectTimeout time.Duration) structs.UpstreamConfig {
+		if cfg.Protocol == "" {
+			cfg.Protocol = chain.Protocol
+		}
+
+		if cfg.Protocol == "" {
+			cfg.Protocol = "tcp"
+		}
+
+		if cfg.ConnectTimeoutMs == 0 {
+			cfg.ConnectTimeoutMs = int(connectTimeout / time.Millisecond)
+		}
+		return cfg
+	}
+
 	var escapeHatchCluster *envoy_cluster_v3.Cluster
 	if !forMeshGateway {
-		if cfg.EnvoyClusterJSON != "" {
+		if rawUpstreamConfig.EnvoyClusterJSON != "" {
 			if chain.Default {
 				// If you haven't done anything to setup the discovery chain, then
 				// you can use the envoy_cluster_json escape hatch.
-				escapeHatchCluster, err = makeClusterFromUserConfig(cfg.EnvoyClusterJSON)
+				escapeHatchCluster, err = makeClusterFromUserConfig(rawUpstreamConfig.EnvoyClusterJSON)
 				if err != nil {
 					return nil, err
 				}
@@ -1006,14 +1026,20 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 
 	var out []*envoy_cluster_v3.Cluster
 	for _, node := range chain.Nodes {
-		if node.Type != structs.DiscoveryGraphNodeTypeResolver {
+		switch {
+		case node == nil:
+			return nil, fmt.Errorf("impossible to process a nil node")
+		case node.Type != structs.DiscoveryGraphNodeTypeResolver:
 			continue
+		case node.Resolver == nil:
+			return nil, fmt.Errorf("impossible to process a non-resolver node")
 		}
 		failover := node.Resolver.Failover
 		// These variables are prefixed with primary to avoid shaddowing bugs.
 		primaryTargetID := node.Resolver.Target
 		primaryTarget := chain.Targets[primaryTargetID]
 		primaryClusterName := CustomizeClusterName(primaryTarget.Name, chain)
+		upstreamConfig := finalizeUpstreamConfig(rawUpstreamConfig, node.Resolver.ConnectTimeout)
 		if forMeshGateway {
 			primaryClusterName = meshGatewayExportedClusterNamePrefix + primaryClusterName
 		}
@@ -1026,22 +1052,38 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 			continue
 		}
 
-		type targetClusterOptions struct {
+		type targetClusterOption struct {
 			targetID    string
 			clusterName string
 		}
 
 		// Construct the information required to make  target clusters. When
 		// failover is configured, create the aggregate cluster.
-		var targetClustersOptions []targetClusterOptions
+		var targetClustersOptions []targetClusterOption
 		if failover != nil && !forMeshGateway {
 			var failoverClusterNames []string
 			for _, tid := range append([]string{primaryTargetID}, failover.Targets...) {
 				target := chain.Targets[tid]
-				clusterName := CustomizeClusterName(target.Name, chain)
+				clusterName := target.Name
+				targetUID := proxycfg.NewUpstreamIDFromTargetID(tid)
+				if targetUID.Peer != "" {
+					tbs, ok := upstreamsSnapshot.UpstreamPeerTrustBundles.Get(targetUID.Peer)
+					// We can't generate cluster on peers without the trust bundle. The
+					// trust bundle should be ready soon.
+					if !ok {
+						s.Logger.Debug("peer trust bundle not ready for discovery chain target",
+							"peer", targetUID.Peer,
+							"target", tid,
+						)
+						continue
+					}
+
+					clusterName = generatePeeredClusterName(targetUID, tbs)
+				}
+				clusterName = CustomizeClusterName(clusterName, chain)
 				clusterName = failoverClusterNamePrefix + clusterName
 
-				targetClustersOptions = append(targetClustersOptions, targetClusterOptions{
+				targetClustersOptions = append(targetClustersOptions, targetClusterOption{
 					targetID:    tid,
 					clusterName: clusterName,
 				})
@@ -1070,7 +1112,7 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 
 			out = append(out, c)
 		} else {
-			targetClustersOptions = append(targetClustersOptions, targetClusterOptions{
+			targetClustersOptions = append(targetClustersOptions, targetClusterOption{
 				targetID:    primaryTargetID,
 				clusterName: primaryClusterName,
 			})
@@ -1089,11 +1131,20 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 				Datacenter: target.Datacenter,
 				Service:    target.Service,
 			}.URI().String()
-			if uid.Peer != "" {
+			targetUID := proxycfg.NewUpstreamIDFromTargetID(targetInfo.targetID)
-				return nil, fmt.Errorf("impossible to get a peer discovery chain")
+			s.Logger.Debug("generating cluster for", "cluster", targetInfo.clusterName)
+			if targetUID.Peer != "" {
+				peerMeta := upstreamsSnapshot.UpstreamPeerMeta(targetUID)
+				upstreamCluster, err := s.makeUpstreamClusterForPeerService(targetUID, upstreamConfig, peerMeta, cfgSnap)
+				if err != nil {
+					continue
+				}
+				// Override the cluster name to include the failover-target~ prefix.
+				upstreamCluster.Name = targetInfo.clusterName
+				out = append(out, upstreamCluster)
+				continue
 			}
 
-			s.Logger.Trace("generating cluster for", "cluster", targetInfo.clusterName)
 			c := &envoy_cluster_v3.Cluster{
 				Name:                 targetInfo.clusterName,
 				AltStatName:          targetInfo.clusterName,
@@ -1114,9 +1165,9 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 				},
 				// TODO(peering): make circuit breakers or outlier detection work?
 				CircuitBreakers: &envoy_cluster_v3.CircuitBreakers{
-					Thresholds: makeThresholdsIfNeeded(cfg.Limits),
+					Thresholds: makeThresholdsIfNeeded(upstreamConfig.Limits),
 				},
-				OutlierDetection: ToOutlierDetection(cfg.PassiveHealthCheck),
+				OutlierDetection: ToOutlierDetection(upstreamConfig.PassiveHealthCheck),
 			}
 
 			var lb *structs.LoadBalancer
@@ -1127,19 +1178,7 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 				return nil, fmt.Errorf("failed to apply load balancer configuration to cluster %q: %v", targetInfo.clusterName, err)
 			}
 
-			var proto string
+			if upstreamConfig.Protocol == "http2" || upstreamConfig.Protocol == "grpc" {
-			if !forMeshGateway {
-				proto = cfg.Protocol
-			}
-			if proto == "" {
-				proto = chain.Protocol
-			}
-
-			if proto == "" {
-				proto = "tcp"
-			}
-
-			if proto == "http2" || proto == "grpc" {
 				if err := s.setHttp2ProtocolOptions(c); err != nil {
 					return nil, err
 				}
@@ -1148,7 +1187,7 @@ func (s *ResourceGenerator) makeUpstreamClustersForDiscoveryChain(
 			configureTLS := true
 			if forMeshGateway {
 				// We only initiate TLS if we're doing an L7 proxy.
-				configureTLS = structs.IsProtocolHTTPLike(proto)
+				configureTLS = structs.IsProtocolHTTPLike(upstreamConfig.Protocol)
 			}
 
 			if configureTLS {
@@ -1221,7 +1260,6 @@ func (s *ResourceGenerator) makeExportedUpstreamClustersForMeshGateway(cfgSnap *
 			proxycfg.NewUpstreamIDFromServiceName(svc),
 			nil,
 			chain,
-			nil,
 			cfgSnap,
 			true,
 		)

agent/xds/clusters_test.go

@@ -169,6 +169,18 @@ func TestClustersFromSnapshot(t *testing.T) {
 				}, nil)
 			},
 		},
+		{
+			name: "custom-passive-healthcheck",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshot(t, func(ns *structs.NodeService) {
+					ns.Proxy.Upstreams[0].Config["passive_health_check"] = map[string]interface{}{
+						"enforcing_consecutive_5xx": float64(80),
+						"max_failures":              float64(5),
+						"interval":                  float64(10),
+					}
+				}, nil)
+			},
+		},
 		{
 			name: "custom-max-inbound-connections",
 			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
@@ -257,6 +269,12 @@ func TestClustersFromSnapshot(t *testing.T) {
 				return proxycfg.TestConfigSnapshotDiscoveryChain(t, "failover", nil, nil)
 			},
 		},
+		{
+			name: "connect-proxy-with-chain-and-failover-to-cluster-peer",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshotDiscoveryChain(t, "failover-to-cluster-peer", nil, nil)
+			},
+		},
 		{
 			name: "connect-proxy-with-tcp-chain-failover-through-remote-gateway",
 			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
@@ -495,6 +513,13 @@ func TestClustersFromSnapshot(t *testing.T) {
 					"failover", nil, nil, nil)
 			},
 		},
+		{
+			name: "ingress-with-chain-and-failover-to-cluster-peer",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshotIngressGateway(t, true, "tcp",
+					"failover-to-cluster-peer", nil, nil, nil)
+			},
+		},
 		{
 			name: "ingress-with-tcp-chain-failover-through-remote-gateway",
 			create: func(t testinf.T) *proxycfg.ConfigSnapshot {

agent/xds/config.go

@@ -27,6 +27,12 @@ type ProxyConfig struct {
 	// Note: This escape hatch is compatible with the discovery chain.
 	PublicListenerJSON string `mapstructure:"envoy_public_listener_json"`
 
+	// ListenerTracingJSON is a complete override ("escape hatch") for the
+	// listeners tracing configuration.
+	//
+	// Note: This escape hatch is compatible with the discovery chain.
+	ListenerTracingJSON string `mapstructure:"envoy_listener_tracing_json"`
+
 	// LocalClusterJSON is a complete override ("escape hatch") for the
 	// local application cluster.
 	//
@@ -168,5 +174,10 @@ func ToOutlierDetection(p *structs.PassiveHealthCheck) *envoy_cluster_v3.Outlier
 	if p.MaxFailures != 0 {
 		od.Consecutive_5Xx = &wrappers.UInt32Value{Value: p.MaxFailures}
 	}
+
+	if p.EnforcingConsecutive5xx != nil {
+		od.EnforcingConsecutive_5Xx = &wrappers.UInt32Value{Value: *p.EnforcingConsecutive5xx}
+	}
+
 	return od
 }

agent/xds/delta.go

@@ -81,6 +81,11 @@ const (
 )
 
 func (s *Server) processDelta(stream ADSDeltaStream, reqCh <-chan *envoy_discovery_v3.DeltaDiscoveryRequest) error {
+	// Handle invalid ACL tokens up-front.
+	if _, err := s.authenticate(stream.Context()); err != nil {
+		return err
+	}
+
 	// Loop state
 	var (
 		cfgSnap     *proxycfg.ConfigSnapshot
@@ -200,7 +205,18 @@ func (s *Server) processDelta(stream ADSDeltaStream, reqCh <-chan *envoy_discove
 				}
 			}
 
-		case cfgSnap = <-stateCh:
+		case cs, ok := <-stateCh:
+			if !ok {
+				// stateCh is closed either when *we* cancel the watch (on-exit via defer)
+				// or by the proxycfg.Manager when an irrecoverable error is encountered
+				// such as the ACL token getting deleted.
+				//
+				// We know for sure that this is the latter case, because in the former we
+				// would've already exited this loop.
+				return status.Error(codes.Aborted, "xDS stream terminated due to an irrecoverable error, please try again")
+			}
+			cfgSnap = cs
+
 			newRes, err := generator.allResourcesFromSnapshot(cfgSnap)
 			if err != nil {
 				return status.Errorf(codes.Unavailable, "failed to generate all xDS resources from the snapshot: %v", err)

agent/xds/endpoints.go

@@ -50,14 +50,19 @@ func (s *ResourceGenerator) endpointsFromSnapshotConnectProxy(cfgSnap *proxycfg.
 			cfgSnap.ConnectProxy.PeerUpstreamEndpoints.Len()+
 			len(cfgSnap.ConnectProxy.WatchedUpstreamEndpoints))
 
-	// NOTE: Any time we skip a chain below we MUST also skip that discovery chain in clusters.go
+	getUpstream := func(uid proxycfg.UpstreamID) (*structs.Upstream, bool) {
-	// so that the sets of endpoints generated matches the sets of clusters.
-	for uid, chain := range cfgSnap.ConnectProxy.DiscoveryChain {
 		upstream := cfgSnap.ConnectProxy.UpstreamConfig[uid]
 
 		explicit := upstream.HasLocalPortOrSocket()
 		implicit := cfgSnap.ConnectProxy.IsImplicitUpstream(uid)
-		if !implicit && !explicit {
+		return upstream, !implicit && !explicit
+	}
+
+	// NOTE: Any time we skip a chain below we MUST also skip that discovery chain in clusters.go
+	// so that the sets of endpoints generated matches the sets of clusters.
+	for uid, chain := range cfgSnap.ConnectProxy.DiscoveryChain {
+		upstream, skip := getUpstream(uid)
+		if skip {
 			// Discovery chain is not associated with a known explicit or implicit upstream so it is skipped.
 			continue
 		}
@@ -70,6 +75,7 @@ func (s *ResourceGenerator) endpointsFromSnapshotConnectProxy(cfgSnap *proxycfg.
 		es, err := s.endpointsFromDiscoveryChain(
 			uid,
 			chain,
+			cfgSnap,
 			cfgSnap.Locality,
 			upstreamConfigMap,
 			cfgSnap.ConnectProxy.WatchedUpstreamEndpoints[uid],
@@ -86,12 +92,9 @@ func (s *ResourceGenerator) endpointsFromSnapshotConnectProxy(cfgSnap *proxycfg.
 	// upstream in clusters.go so that the sets of endpoints generated matches
 	// the sets of clusters.
 	for _, uid := range cfgSnap.ConnectProxy.PeeredUpstreamIDs() {
-		upstreamCfg := cfgSnap.ConnectProxy.UpstreamConfig[uid]
+		_, skip := getUpstream(uid)
-
+		if skip {
-		explicit := upstreamCfg.HasLocalPortOrSocket()
+			// Discovery chain is not associated with a known explicit or implicit upstream so it is skipped.
-		implicit := cfgSnap.ConnectProxy.IsImplicitUpstream(uid)
-		if !implicit && !explicit {
-			// Not associated with a known explicit or implicit upstream so it is skipped.
 			continue
 		}
 
@@ -104,22 +107,14 @@ func (s *ResourceGenerator) endpointsFromSnapshotConnectProxy(cfgSnap *proxycfg.
 
 		clusterName := generatePeeredClusterName(uid, tbs)
 
-		// Also skip peer instances with a hostname as their address. EDS
+		loadAssignment, err := s.makeUpstreamLoadAssignmentForPeerService(cfgSnap, clusterName, uid)
-		// cannot resolve hostnames, so we provide them through CDS instead.
+
-		if _, ok := cfgSnap.ConnectProxy.PeerUpstreamEndpointsUseHostnames[uid]; ok {
+		if err != nil {
-			continue
+			return nil, err
 		}
 
-		endpoints, ok := cfgSnap.ConnectProxy.PeerUpstreamEndpoints.Get(uid)
+		if loadAssignment != nil {
-		if ok {
+			resources = append(resources, loadAssignment)
-			la := makeLoadAssignment(
-				clusterName,
-				[]loadAssignmentEndpointGroup{
-					{Endpoints: endpoints},
-				},
-				proxycfg.GatewayKey{ /*empty so it never matches*/ },
-			)
-			resources = append(resources, la)
 		}
 	}
 
@@ -375,6 +370,7 @@ func (s *ResourceGenerator) endpointsFromSnapshotIngressGateway(cfgSnap *proxycf
 			es, err := s.endpointsFromDiscoveryChain(
 				uid,
 				cfgSnap.IngressGateway.DiscoveryChain[uid],
+				cfgSnap,
 				proxycfg.GatewayKey{Datacenter: cfgSnap.Datacenter, Partition: u.DestinationPartition},
 				u.Config,
 				cfgSnap.IngressGateway.WatchedUpstreamEndpoints[uid],
@@ -412,9 +408,38 @@ func makePipeEndpoint(path string) *envoy_endpoint_v3.LbEndpoint {
 	}
 }
 
+func (s *ResourceGenerator) makeUpstreamLoadAssignmentForPeerService(cfgSnap *proxycfg.ConfigSnapshot, clusterName string, uid proxycfg.UpstreamID) (*envoy_endpoint_v3.ClusterLoadAssignment, error) {
+	var la *envoy_endpoint_v3.ClusterLoadAssignment
+
+	upstreamsSnapshot, err := cfgSnap.ToConfigSnapshotUpstreams()
+	if err != nil {
+		return la, err
+	}
+
+	// Also skip peer instances with a hostname as their address. EDS
+	// cannot resolve hostnames, so we provide them through CDS instead.
+	if _, ok := upstreamsSnapshot.PeerUpstreamEndpointsUseHostnames[uid]; ok {
+		return la, nil
+	}
+
+	endpoints, ok := upstreamsSnapshot.PeerUpstreamEndpoints.Get(uid)
+	if !ok {
+		return nil, nil
+	}
+	la = makeLoadAssignment(
+		clusterName,
+		[]loadAssignmentEndpointGroup{
+			{Endpoints: endpoints},
+		},
+		proxycfg.GatewayKey{ /*empty so it never matches*/ },
+	)
+	return la, nil
+}
+
 func (s *ResourceGenerator) endpointsFromDiscoveryChain(
 	uid proxycfg.UpstreamID,
 	chain *structs.CompiledDiscoveryChain,
+	cfgSnap *proxycfg.ConfigSnapshot,
 	gatewayKey proxycfg.GatewayKey,
 	upstreamConfigMap map[string]interface{},
 	upstreamEndpoints map[string]structs.CheckServiceNodes,
@@ -432,6 +457,14 @@ func (s *ResourceGenerator) endpointsFromDiscoveryChain(
 		upstreamConfigMap = make(map[string]interface{}) // TODO:needed?
 	}
 
+	upstreamsSnapshot, err := cfgSnap.ToConfigSnapshotUpstreams()
+
+	// Mesh gateways are exempt because upstreamsSnapshot is only used for
+	// cluster peering targets and transative failover/redirects are unsupported.
+	if err != nil && !forMeshGateway {
+		return nil, fmt.Errorf("No upstream snapshot for gateway mode %q", cfgSnap.Kind)
+	}
+
 	var resources []proto.Message
 
 	var escapeHatchCluster *envoy_cluster_v3.Cluster
@@ -465,8 +498,15 @@ func (s *ResourceGenerator) endpointsFromDiscoveryChain(
 		if node.Type != structs.DiscoveryGraphNodeTypeResolver {
 			continue
 		}
+		primaryTargetID := node.Resolver.Target
 		failover := node.Resolver.Failover
 
+		type targetLoadAssignmentOption struct {
+			targetID    string
+			clusterName string
+		}
+		var targetLoadAssignmentOptions []targetLoadAssignmentOption
+
 		var numFailoverTargets int
 		if failover != nil {
 			numFailoverTargets = len(failover.Targets)
@@ -474,66 +514,84 @@ func (s *ResourceGenerator) endpointsFromDiscoveryChain(
 		clusterNamePrefix := ""
 		if numFailoverTargets > 0 && !forMeshGateway {
 			clusterNamePrefix = failoverClusterNamePrefix
-			for _, failTargetID := range failover.Targets {
+			for _, targetID := range append([]string{primaryTargetID}, failover.Targets...) {
-				target := chain.Targets[failTargetID]
+				target := chain.Targets[targetID]
-				endpointGroup, valid := makeLoadAssignmentEndpointGroup(
+				clusterName := target.Name
-					chain.Targets,
+				targetUID := proxycfg.NewUpstreamIDFromTargetID(targetID)
-					upstreamEndpoints,
+				if targetUID.Peer != "" {
-					gatewayEndpoints,
+					tbs, ok := upstreamsSnapshot.UpstreamPeerTrustBundles.Get(targetUID.Peer)
-					failTargetID,
+					// We can't generate cluster on peers without the trust bundle. The
-					gatewayKey,
+					// trust bundle should be ready soon.
-					forMeshGateway,
+					if !ok {
-				)
+						s.Logger.Debug("peer trust bundle not ready for discovery chain target",
-				if !valid {
+							"peer", targetUID.Peer,
-					continue // skip the failover target if we're still populating the snapshot
+							"target", targetID,
-				}
+						)
+						continue
+					}
 
-				clusterName := CustomizeClusterName(target.Name, chain)
+					clusterName = generatePeeredClusterName(targetUID, tbs)
+				}
+				clusterName = CustomizeClusterName(clusterName, chain)
 				clusterName = failoverClusterNamePrefix + clusterName
 				if escapeHatchCluster != nil {
 					clusterName = escapeHatchCluster.Name
 				}
 
-				s.Logger.Debug("generating endpoints for", "cluster", clusterName)
+				targetLoadAssignmentOptions = append(targetLoadAssignmentOptions, targetLoadAssignmentOption{
-
+					targetID:    targetID,
-				la := makeLoadAssignment(
+					clusterName: clusterName,
-					clusterName,
+				})
-					[]loadAssignmentEndpointGroup{endpointGroup},
-					gatewayKey,
-				)
-				resources = append(resources, la)
 			}
-		}
+		} else {
-		targetID := node.Resolver.Target
+			target := chain.Targets[primaryTargetID]
-
+			clusterName := CustomizeClusterName(target.Name, chain)
-		target := chain.Targets[targetID]
+			clusterName = clusterNamePrefix + clusterName
-		clusterName := CustomizeClusterName(target.Name, chain)
+			if escapeHatchCluster != nil {
-		clusterName = clusterNamePrefix + clusterName
+				clusterName = escapeHatchCluster.Name
-		if escapeHatchCluster != nil {
+			}
-			clusterName = escapeHatchCluster.Name
+			if forMeshGateway {
-		}
+				clusterName = meshGatewayExportedClusterNamePrefix + clusterName
-		if forMeshGateway {
+			}
-			clusterName = meshGatewayExportedClusterNamePrefix + clusterName
+			targetLoadAssignmentOptions = append(targetLoadAssignmentOptions, targetLoadAssignmentOption{
-		}
+				targetID:    primaryTargetID,
-		s.Logger.Debug("generating endpoints for", "cluster", clusterName)
+				clusterName: clusterName,
-		endpointGroup, valid := makeLoadAssignmentEndpointGroup(
+			})
-			chain.Targets,
-			upstreamEndpoints,
-			gatewayEndpoints,
-			targetID,
-			gatewayKey,
-			forMeshGateway,
-		)
-		if !valid {
-			continue // skip the cluster if we're still populating the snapshot
 		}
 
-		la := makeLoadAssignment(
+		for _, targetInfo := range targetLoadAssignmentOptions {
-			clusterName,
+			s.Logger.Debug("generating endpoints for", "cluster", targetInfo.clusterName)
-			[]loadAssignmentEndpointGroup{endpointGroup},
+			targetUID := proxycfg.NewUpstreamIDFromTargetID(targetInfo.targetID)
-			gatewayKey,
+			if targetUID.Peer != "" {
-		)
+				loadAssignment, err := s.makeUpstreamLoadAssignmentForPeerService(cfgSnap, targetInfo.clusterName, targetUID)
-		resources = append(resources, la)
+				if err != nil {
+					return nil, err
+				}
+				if loadAssignment != nil {
+					resources = append(resources, loadAssignment)
+				}
+				continue
+			}
+
+			endpointGroup, valid := makeLoadAssignmentEndpointGroup(
+				chain.Targets,
+				upstreamEndpoints,
+				gatewayEndpoints,
+				targetInfo.targetID,
+				gatewayKey,
+				forMeshGateway,
+			)
+			if !valid {
+				continue // skip the cluster if we're still populating the snapshot
+			}
+
+			la := makeLoadAssignment(
+				targetInfo.clusterName,
+				[]loadAssignmentEndpointGroup{endpointGroup},
+				gatewayKey,
+			)
+			resources = append(resources, la)
+		}
 	}
 
 	return resources, nil
@@ -586,6 +644,7 @@ func (s *ResourceGenerator) makeExportedUpstreamEndpointsForMeshGateway(cfgSnap
 		clusterEndpoints, err := s.endpointsFromDiscoveryChain(
 			proxycfg.NewUpstreamIDFromServiceName(svc),
 			chain,
+			cfgSnap,
 			cfgSnap.Locality,
 			nil,
 			chainEndpoints,
@@ -640,11 +699,12 @@ func makeLoadAssignment(clusterName string, endpointGroups []loadAssignmentEndpo
 				healthStatus = endpointGroup.OverrideHealth
 			}
 
+			endpoint := &envoy_endpoint_v3.Endpoint{
+				Address: makeAddress(addr, port),
+			}
 			es = append(es, &envoy_endpoint_v3.LbEndpoint{
 				HostIdentifier: &envoy_endpoint_v3.LbEndpoint_Endpoint{
-					Endpoint: &envoy_endpoint_v3.Endpoint{
+					Endpoint: endpoint,
-						Address: makeAddress(addr, port),
-					},
 				},
 				HealthStatus:        healthStatus,
 				LoadBalancingWeight: makeUint32Value(weight),

agent/xds/endpoints_test.go

@@ -284,6 +284,12 @@ func TestEndpointsFromSnapshot(t *testing.T) {
 				return proxycfg.TestConfigSnapshotDiscoveryChain(t, "failover", nil, nil)
 			},
 		},
+		{
+			name: "connect-proxy-with-chain-and-failover-to-cluster-peer",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshotDiscoveryChain(t, "failover-to-cluster-peer", nil, nil)
+			},
+		},
 		{
 			name: "connect-proxy-with-tcp-chain-failover-through-remote-gateway",
 			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
@@ -396,6 +402,13 @@ func TestEndpointsFromSnapshot(t *testing.T) {
 					"failover", nil, nil, nil)
 			},
 		},
+		{
+			name: "ingress-with-chain-and-failover-to-cluster-peer",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshotIngressGateway(t, true, "tcp",
+					"failover-to-cluster-peer", nil, nil, nil)
+			},
+		},
 		{
 			name: "ingress-with-tcp-chain-failover-through-remote-gateway",
 			create: func(t testinf.T) *proxycfg.ConfigSnapshot {

agent/xds/failover_math_test.go

@@ -15,15 +15,40 @@ func TestFirstHealthyTarget(t *testing.T) {
 	warning := proxycfg.TestUpstreamNodesInStatus(t, "warning")
 	critical := proxycfg.TestUpstreamNodesInStatus(t, "critical")
 
-	warnOnlyPassingTarget := structs.NewDiscoveryTarget("all-warn", "", "default", "default", "dc1")
+	warnOnlyPassingTarget := structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+		Service:    "all-warn",
+		Namespace:  "default",
+		Partition:  "default",
+		Datacenter: "dc1",
+	})
 	warnOnlyPassingTarget.Subset.OnlyPassing = true
-	failOnlyPassingTarget := structs.NewDiscoveryTarget("all-fail", "", "default", "default", "dc1")
+	failOnlyPassingTarget := structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+		Service:    "all-fail",
+		Namespace:  "default",
+		Partition:  "default",
+		Datacenter: "dc1",
+	})
 	failOnlyPassingTarget.Subset.OnlyPassing = true
 
 	targets := map[string]*structs.DiscoveryTarget{
-		"all-ok.default.dc1":               structs.NewDiscoveryTarget("all-ok", "", "default", "default", "dc1"),
+		"all-ok.default.dc1": structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
-		"all-warn.default.dc1":             structs.NewDiscoveryTarget("all-warn", "", "default", "default", "dc1"),
+			Service:    "all-ok",
-		"all-fail.default.default.dc1":     structs.NewDiscoveryTarget("all-fail", "", "default", "default", "dc1"),
+			Namespace:  "default",
+			Partition:  "default",
+			Datacenter: "dc1",
+		}),
+		"all-warn.default.dc1": structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:    "all-warn",
+			Namespace:  "default",
+			Partition:  "default",
+			Datacenter: "dc1",
+		}),
+		"all-fail.default.default.dc1": structs.NewDiscoveryTarget(structs.DiscoveryTargetOpts{
+			Service:    "all-fail",
+			Namespace:  "default",
+			Partition:  "default",
+			Datacenter: "dc1",
+		}),
 		"all-warn-onlypassing.default.dc1": warnOnlyPassingTarget,
 		"all-fail-onlypassing.default.dc1": failOnlyPassingTarget,
 	}

agent/xds/listeners.go

@@ -3,7 +3,6 @@ package xds
 import (
 	"errors"
 	"fmt"
-	envoy_extensions_filters_listener_http_inspector_v3 "github.com/envoyproxy/go-control-plane/envoy/extensions/filters/listener/http_inspector/v3"
 	"net"
 	"net/url"
 	"regexp"
@@ -12,6 +11,8 @@ import (
 	"strings"
 	"time"
 
+	envoy_extensions_filters_listener_http_inspector_v3 "github.com/envoyproxy/go-control-plane/envoy/extensions/filters/listener/http_inspector/v3"
+
 	envoy_core_v3 "github.com/envoyproxy/go-control-plane/envoy/config/core/v3"
 	envoy_listener_v3 "github.com/envoyproxy/go-control-plane/envoy/config/listener/v3"
 	envoy_route_v3 "github.com/envoyproxy/go-control-plane/envoy/config/route/v3"
@@ -107,6 +108,19 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 		}
 	}
 
+	proxyCfg, err := ParseProxyConfig(cfgSnap.Proxy.Config)
+	if err != nil {
+		// Don't hard fail on a config typo, just warn. The parse func returns
+		// default config if there is an error so it's safe to continue.
+		s.Logger.Warn("failed to parse Connect.Proxy.Config", "error", err)
+	}
+	var tracing *envoy_http_v3.HttpConnectionManager_Tracing
+	if proxyCfg.ListenerTracingJSON != "" {
+		if tracing, err = makeTracingFromUserConfig(proxyCfg.ListenerTracingJSON); err != nil {
+			s.Logger.Warn("failed to parse ListenerTracingJSON config", "error", err)
+		}
+	}
+
 	for uid, chain := range cfgSnap.ConnectProxy.DiscoveryChain {
 		upstreamCfg := cfgSnap.ConnectProxy.UpstreamConfig[uid]
 
@@ -153,6 +167,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 				filterName:  filterName,
 				protocol:    cfg.Protocol,
 				useRDS:      useRDS,
+				tracing:     tracing,
 			})
 			if err != nil {
 				return nil, err
@@ -178,6 +193,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 			filterName:  filterName,
 			protocol:    cfg.Protocol,
 			useRDS:      useRDS,
+			tracing:     tracing,
 		})
 		if err != nil {
 			return nil, err
@@ -249,6 +265,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 				filterName: routeName,
 				protocol:   svcConfig.Protocol,
 				useRDS:     true,
+				tracing:    tracing,
 			})
 			if err != nil {
 				return err
@@ -265,6 +282,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 					clusterName: clusterName,
 					filterName:  clusterName,
 					protocol:    svcConfig.Protocol,
+					tracing:     tracing,
 				})
 				if err != nil {
 					return err
@@ -376,6 +394,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 			protocol:   cfg.Protocol,
 			useRDS:     false,
 			statPrefix: "upstream_peered.",
+			tracing:    tracing,
 		})
 		if err != nil {
 			return nil, err
@@ -533,6 +552,7 @@ func (s *ResourceGenerator) listenersFromSnapshotConnectProxy(cfgSnap *proxycfg.
 			filterName:  uid.EnvoyID(),
 			routeName:   uid.EnvoyID(),
 			protocol:    cfg.Protocol,
+			tracing:     tracing,
 		})
 		if err != nil {
 			return nil, err
@@ -1188,12 +1208,20 @@ func (s *ResourceGenerator) makeInboundListener(cfgSnap *proxycfg.ConfigSnapshot
 
 	l = makePortListener(name, addr, port, envoy_core_v3.TrafficDirection_INBOUND)
 
+	var tracing *envoy_http_v3.HttpConnectionManager_Tracing
+	if cfg.ListenerTracingJSON != "" {
+		if tracing, err = makeTracingFromUserConfig(cfg.ListenerTracingJSON); err != nil {
+			s.Logger.Warn("failed to parse ListenerTracingJSON config", "error", err)
+		}
+	}
+
 	filterOpts := listenerFilterOpts{
 		protocol:         cfg.Protocol,
 		filterName:       name,
 		routeName:        name,
 		cluster:          LocalAppClusterName,
 		requestTimeoutMs: cfg.LocalRequestTimeoutMs,
+		tracing:          tracing,
 	}
 	if useHTTPFilter {
 		filterOpts.httpAuthzFilter, err = makeRBACHTTPFilter(
@@ -1214,16 +1242,38 @@ func (s *ResourceGenerator) makeInboundListener(cfgSnap *proxycfg.ConfigSnapshot
 			filterOpts.forwardClientPolicy = envoy_http_v3.HttpConnectionManager_APPEND_FORWARD
 		}
 	}
+
+	// If an inbound connect limit is set, inject a connection limit filter on each chain.
+	if cfg.MaxInboundConnections > 0 {
+		connectionLimitFilter, err := makeConnectionLimitFilter(cfg.MaxInboundConnections)
+		if err != nil {
+			return nil, err
+		}
+		l.FilterChains = []*envoy_listener_v3.FilterChain{
+			{
+				Filters: []*envoy_listener_v3.Filter{
+					connectionLimitFilter,
+				},
+			},
+		}
+	}
+
 	filter, err := makeListenerFilter(filterOpts)
 	if err != nil {
 		return nil, err
 	}
-	l.FilterChains = []*envoy_listener_v3.FilterChain{
+
-		{
+	if len(l.FilterChains) > 0 {
-			Filters: []*envoy_listener_v3.Filter{
+		// The list of FilterChains has already been initialized
-				filter,
+		l.FilterChains[0].Filters = append(l.FilterChains[0].Filters, filter)
+	} else {
+		l.FilterChains = []*envoy_listener_v3.FilterChain{
+			{
+				Filters: []*envoy_listener_v3.Filter{
+					filter,
+				},
 			},
-		},
+		}
 	}
 
 	err = s.finalizePublicListenerFromConfig(l, cfgSnap, cfg, useHTTPFilter)
@@ -1249,17 +1299,6 @@ func (s *ResourceGenerator) finalizePublicListenerFromConfig(l *envoy_listener_v
 		return nil
 	}
 
-	// If an inbound connect limit is set, inject a connection limit filter on each chain.
-	if proxyCfg.MaxInboundConnections > 0 {
-		filter, err := makeConnectionLimitFilter(proxyCfg.MaxInboundConnections)
-		if err != nil {
-			return nil
-		}
-		for idx := range l.FilterChains {
-			l.FilterChains[idx].Filters = append(l.FilterChains[idx].Filters, filter)
-		}
-	}
-
 	return nil
 }
 
@@ -1299,6 +1338,7 @@ func (s *ResourceGenerator) makeExposedCheckListener(cfgSnap *proxycfg.ConfigSna
 		statPrefix:      "",
 		routePath:       path.Path,
 		httpAuthzFilter: nil,
+		// in the exposed check listener we don't set the tracing configuration
 	}
 	f, err := makeListenerFilter(opts)
 	if err != nil {
@@ -1531,6 +1571,19 @@ func (s *ResourceGenerator) makeFilterChainTerminatingGateway(cfgSnap *proxycfg.
 		filterChain.Filters = append(filterChain.Filters, authFilter)
 	}
 
+	proxyCfg, err := ParseProxyConfig(cfgSnap.Proxy.Config)
+	if err != nil {
+		// Don't hard fail on a config typo, just warn. The parse func returns
+		// default config if there is an error so it's safe to continue.
+		s.Logger.Warn("failed to parse Connect.Proxy.Config", "error", err)
+	}
+	var tracing *envoy_http_v3.HttpConnectionManager_Tracing
+	if proxyCfg.ListenerTracingJSON != "" {
+		if tracing, err = makeTracingFromUserConfig(proxyCfg.ListenerTracingJSON); err != nil {
+			s.Logger.Warn("failed to parse ListenerTracingJSON config", "error", err)
+		}
+	}
+
 	// Lastly we setup the actual proxying component. For L4 this is a straight
 	// tcp proxy. For L7 this is a very hands-off HTTP proxy just to inject an
 	// HTTP filter to do intention checks here instead.
@@ -1541,6 +1594,7 @@ func (s *ResourceGenerator) makeFilterChainTerminatingGateway(cfgSnap *proxycfg.
 		cluster:    tgtwyOpts.cluster,
 		statPrefix: "upstream.",
 		routePath:  "",
+		tracing:    tracing,
 	}
 
 	if useHTTPFilter {
@@ -1787,6 +1841,7 @@ type filterChainOpts struct {
 	statPrefix           string
 	forwardClientDetails bool
 	forwardClientPolicy  envoy_http_v3.HttpConnectionManager_ForwardClientCertDetails
+	tracing              *envoy_http_v3.HttpConnectionManager_Tracing
 }
 
 func (s *ResourceGenerator) makeUpstreamFilterChain(opts filterChainOpts) (*envoy_listener_v3.FilterChain, error) {
@@ -1802,6 +1857,7 @@ func (s *ResourceGenerator) makeUpstreamFilterChain(opts filterChainOpts) (*envo
 		statPrefix:           opts.statPrefix,
 		forwardClientDetails: opts.forwardClientDetails,
 		forwardClientPolicy:  opts.forwardClientPolicy,
+		tracing:              opts.tracing,
 	})
 	if err != nil {
 		return nil, err
@@ -1944,6 +2000,7 @@ type listenerFilterOpts struct {
 	httpAuthzFilter      *envoy_http_v3.HttpFilter
 	forwardClientDetails bool
 	forwardClientPolicy  envoy_http_v3.HttpConnectionManager_ForwardClientCertDetails
+	tracing              *envoy_http_v3.HttpConnectionManager_Tracing
 }
 
 func makeListenerFilter(opts listenerFilterOpts) (*envoy_listener_v3.Filter, error) {
@@ -1990,6 +2047,7 @@ func makeTCPProxyFilter(filterName, cluster, statPrefix string) (*envoy_listener
 
 func makeConnectionLimitFilter(limit int) (*envoy_listener_v3.Filter, error) {
 	cfg := &envoy_connection_limit_v3.ConnectionLimit{
+		StatPrefix:     "inbound_connection_limit",
 		MaxConnections: wrapperspb.UInt64(uint64(limit)),
 	}
 	return makeFilter("envoy.filters.network.connection_limit", cfg)
@@ -2002,6 +2060,19 @@ func makeStatPrefix(prefix, filterName string) string {
 	return fmt.Sprintf("%s%s", prefix, strings.Replace(filterName, ":", "_", -1))
 }
 
+func makeTracingFromUserConfig(configJSON string) (*envoy_http_v3.HttpConnectionManager_Tracing, error) {
+	// Type field is present so decode it as a any.Any
+	var any any.Any
+	if err := jsonpb.UnmarshalString(configJSON, &any); err != nil {
+		return nil, err
+	}
+	var t envoy_http_v3.HttpConnectionManager_Tracing
+	if err := proto.Unmarshal(any.Value, &t); err != nil {
+		return nil, err
+	}
+	return &t, nil
+}
+
 func makeHTTPFilter(opts listenerFilterOpts) (*envoy_listener_v3.Filter, error) {
 	router, err := makeEnvoyHTTPFilter("envoy.filters.http.router", &envoy_http_router_v3.Router{})
 	if err != nil {
@@ -2022,6 +2093,10 @@ func makeHTTPFilter(opts listenerFilterOpts) (*envoy_listener_v3.Filter, error)
 		},
 	}
 
+	if opts.tracing != nil {
+		cfg.Tracing = opts.tracing
+	}
+
 	if opts.useRDS {
 		if opts.cluster != "" {
 			return nil, fmt.Errorf("cannot specify cluster name when using RDS")

agent/xds/listeners_test.go

@@ -772,6 +772,15 @@ func TestListenersFromSnapshot(t *testing.T) {
 			name:   "transparent-proxy-terminating-gateway",
 			create: proxycfg.TestConfigSnapshotTransparentProxyTerminatingGatewayCatalogDestinationsOnly,
 		},
+		{
+			name: "custom-trace-listener",
+			create: func(t testinf.T) *proxycfg.ConfigSnapshot {
+				return proxycfg.TestConfigSnapshot(t, func(ns *structs.NodeService) {
+					ns.Proxy.Config["protocol"] = "http"
+					ns.Proxy.Config["envoy_listener_tracing_json"] = customTraceJSON(t)
+				}, nil)
+			},
+		},
 	}
 
 	latestEnvoyVersion := proxysupport.EnvoyVersions[0]
@@ -947,6 +956,40 @@ func customHTTPListenerJSON(t testinf.T, opts customHTTPListenerJSONOptions) str
 	return buf.String()
 }
 
+func customTraceJSON(t testinf.T) string {
+	t.Helper()
+	return `
+	{
+        "@type" : "type.googleapis.com/envoy.extensions.filters.network.http_connection_manager.v3.HttpConnectionManager.Tracing",
+        "provider" : {
+          "name" : "envoy.tracers.zipkin",
+          "typed_config" : {
+            "@type" : "type.googleapis.com/envoy.config.trace.v3.ZipkinConfig",
+            "collector_cluster" : "otelcolector",
+            "collector_endpoint" : "/api/v2/spans",
+            "collector_endpoint_version" : "HTTP_JSON",
+            "shared_span_context" : false
+          }
+        },
+        "custom_tags" : [
+          {
+            "tag" : "custom_header",
+            "request_header" : {
+              "name" : "x-custom-traceid",
+              "default_value" : ""
+            }
+          },
+          {
+            "tag" : "alloc_id",
+            "environment" : {
+              "name" : "NOMAD_ALLOC_ID"
+            }
+          }
+        ]
+      }
+	`
+}
+
 type configFetcherFunc func() string
 
 var _ ConfigFetcher = (configFetcherFunc)(nil)

agent/xds/server.go

@@ -186,6 +186,18 @@ func (s *Server) Register(srv *grpc.Server) {
 	envoy_discovery_v3.RegisterAggregatedDiscoveryServiceServer(srv, s)
 }
 
+func (s *Server) authenticate(ctx context.Context) (acl.Authorizer, error) {
+	authz, err := s.ResolveToken(external.TokenFromContext(ctx))
+	if acl.IsErrNotFound(err) {
+		return nil, status.Errorf(codes.Unauthenticated, "unauthenticated: %v", err)
+	} else if acl.IsErrPermissionDenied(err) {
+		return nil, status.Error(codes.PermissionDenied, err.Error())
+	} else if err != nil {
+		return nil, status.Errorf(codes.Internal, "error resolving acl token: %v", err)
+	}
+	return authz, nil
+}
+
 // authorize the xDS request using the token stored in ctx. This authorization is
 // a bit different from most interfaces. Instead of explicitly authorizing or
 // filtering each piece of data in the response, the request is authorized
@@ -201,13 +213,9 @@ func (s *Server) authorize(ctx context.Context, cfgSnap *proxycfg.ConfigSnapshot
 		return status.Errorf(codes.Unauthenticated, "unauthenticated: no config snapshot")
 	}
 
-	authz, err := s.ResolveToken(external.TokenFromContext(ctx))
+	authz, err := s.authenticate(ctx)
-	if acl.IsErrNotFound(err) {
+	if err != nil {
-		return status.Errorf(codes.Unauthenticated, "unauthenticated: %v", err)
+		return err
-	} else if acl.IsErrPermissionDenied(err) {
-		return status.Error(codes.PermissionDenied, err.Error())
-	} else if err != nil {
-		return status.Errorf(codes.Internal, "error resolving acl token: %v", err)
 	}
 
 	var authzContext acl.AuthorizerContext