consul/agent/proxycfg/testing.go

1209 lines
35 KiB
Go
Raw Normal View History

package proxycfg
import (
"context"
"fmt"
"path"
"sync"
"sync/atomic"
"time"
"github.com/hashicorp/consul/agent/cache"
cachetype "github.com/hashicorp/consul/agent/cache-types"
"github.com/hashicorp/consul/agent/connect"
"github.com/hashicorp/consul/agent/consul/discoverychain"
"github.com/hashicorp/consul/agent/structs"
"github.com/mitchellh/go-testing-interface"
"github.com/stretchr/testify/require"
)
// TestCacheTypes encapsulates all the different cache types proxycfg.State will
// watch/request for controlling one during testing.
type TestCacheTypes struct {
roots *ControllableCacheType
leaf *ControllableCacheType
intentions *ControllableCacheType
health *ControllableCacheType
query *ControllableCacheType
compiledChain *ControllableCacheType
serviceHTTPChecks *ControllableCacheType
}
// NewTestCacheTypes creates a set of ControllableCacheTypes for all types that
// proxycfg will watch suitable for testing a proxycfg.State or Manager.
func NewTestCacheTypes(t testing.T) *TestCacheTypes {
t.Helper()
ct := &TestCacheTypes{
roots: NewControllableCacheType(t),
leaf: NewControllableCacheType(t),
intentions: NewControllableCacheType(t),
health: NewControllableCacheType(t),
query: NewControllableCacheType(t),
compiledChain: NewControllableCacheType(t),
serviceHTTPChecks: NewControllableCacheType(t),
}
ct.query.blocking = false
return ct
}
// TestCacheWithTypes registers ControllableCacheTypes for all types that
// proxycfg will watch suitable for testing a proxycfg.State or Manager.
func TestCacheWithTypes(t testing.T, types *TestCacheTypes) *cache.Cache {
c := cache.TestCache(t)
c.RegisterType(cachetype.ConnectCARootName, types.roots)
c.RegisterType(cachetype.ConnectCALeafName, types.leaf)
c.RegisterType(cachetype.IntentionMatchName, types.intentions)
c.RegisterType(cachetype.HealthServicesName, types.health)
c.RegisterType(cachetype.PreparedQueryName, types.query)
c.RegisterType(cachetype.CompiledDiscoveryChainName, types.compiledChain)
c.RegisterType(cachetype.ServiceHTTPChecksName, types.serviceHTTPChecks)
return c
}
// TestCerts generates a CA and Leaf suitable for returning as mock CA
// root/leaf cache requests.
func TestCerts(t testing.T) (*structs.IndexedCARoots, *structs.IssuedCert) {
t.Helper()
ca := connect.TestCA(t, nil)
roots := &structs.IndexedCARoots{
ActiveRootID: ca.ID,
TrustDomain: fmt.Sprintf("%s.consul", connect.TestClusterID),
Roots: []*structs.CARoot{ca},
}
return roots, TestLeafForCA(t, ca)
}
// TestLeafForCA generates new Leaf suitable for returning as mock CA
// leaf cache response, signed by an existing CA.
func TestLeafForCA(t testing.T, ca *structs.CARoot) *structs.IssuedCert {
leafPEM, pkPEM := connect.TestLeaf(t, "web", ca)
leafCert, err := connect.ParseCert(leafPEM)
require.NoError(t, err)
return &structs.IssuedCert{
SerialNumber: connect.EncodeSerialNumber(leafCert.SerialNumber),
CertPEM: leafPEM,
PrivateKeyPEM: pkPEM,
Service: "web",
ServiceURI: leafCert.URIs[0].String(),
ValidAfter: leafCert.NotBefore,
ValidBefore: leafCert.NotAfter,
}
}
// TestIntentions returns a sample intentions match result useful to
// mocking service discovery cache results.
func TestIntentions(t testing.T) *structs.IndexedIntentionMatches {
return &structs.IndexedIntentionMatches{
Matches: []structs.Intentions{
[]*structs.Intention{
&structs.Intention{
ID: "foo",
SourceNS: "default",
SourceName: "billing",
DestinationNS: "default",
DestinationName: "web",
Action: structs.IntentionActionAllow,
},
},
},
}
}
// TestUpstreamNodes returns a sample service discovery result useful to
// mocking service discovery cache results.
func TestUpstreamNodes(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test1",
Node: "test1",
Address: "10.10.1.1",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test2",
Node: "test2",
Address: "10.10.1.2",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
},
}
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
func TestUpstreamNodesInStatus(t testing.T, status string) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test1",
Node: "test1",
Address: "10.10.1.1",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
Checks: structs.HealthChecks{
&structs.HealthCheck{
Node: "test1",
ServiceName: "web",
Name: "force",
Status: status,
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test2",
Node: "test2",
Address: "10.10.1.2",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
Checks: structs.HealthChecks{
&structs.HealthCheck{
Node: "test2",
ServiceName: "web",
Name: "force",
Status: status,
},
},
},
}
}
func TestUpstreamNodesDC2(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test1",
Node: "test1",
Address: "10.20.1.1",
Datacenter: "dc2",
},
Service: structs.TestNodeService(t),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test2",
Node: "test2",
Address: "10.20.1.2",
Datacenter: "dc2",
},
Service: structs.TestNodeService(t),
},
}
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
func TestUpstreamNodesInStatusDC2(t testing.T, status string) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test1",
Node: "test1",
Address: "10.20.1.1",
Datacenter: "dc2",
},
Service: structs.TestNodeService(t),
Checks: structs.HealthChecks{
&structs.HealthCheck{
Node: "test1",
ServiceName: "web",
Name: "force",
Status: status,
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test2",
Node: "test2",
Address: "10.20.1.2",
Datacenter: "dc2",
},
Service: structs.TestNodeService(t),
Checks: structs.HealthChecks{
&structs.HealthCheck{
Node: "test2",
ServiceName: "web",
Name: "force",
Status: status,
},
},
},
}
}
func TestUpstreamNodesDC3(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test1",
Node: "test1",
Address: "10.30.1.1",
Datacenter: "dc3",
},
Service: structs.TestNodeService(t),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "test2",
Node: "test2",
Address: "10.30.1.2",
Datacenter: "dc3",
},
Service: structs.TestNodeService(t),
},
}
}
func TestUpstreamNodesAlternate(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "alt-test1",
Node: "alt-test1",
Address: "10.20.1.1",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "alt-test2",
Node: "alt-test2",
Address: "10.20.1.2",
Datacenter: "dc1",
},
Service: structs.TestNodeService(t),
},
}
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
func TestGatewayNodesDC1(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-1",
Node: "mesh-gateway",
Address: "10.10.1.1",
Datacenter: "dc1",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.10.1.1", 8443,
structs.ServiceAddress{Address: "10.10.1.1", Port: 8443},
structs.ServiceAddress{Address: "198.118.1.1", Port: 443}),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-2",
Node: "mesh-gateway",
Address: "10.10.1.2",
Datacenter: "dc1",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.10.1.2", 8443,
structs.ServiceAddress{Address: "10.0.1.2", Port: 8443},
structs.ServiceAddress{Address: "198.118.1.2", Port: 443}),
},
}
}
func TestGatewayNodesDC2(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-1",
Node: "mesh-gateway",
Address: "10.0.1.1",
Datacenter: "dc2",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.0.1.1", 8443,
structs.ServiceAddress{Address: "10.0.1.1", Port: 8443},
structs.ServiceAddress{Address: "198.18.1.1", Port: 443}),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-2",
Node: "mesh-gateway",
Address: "10.0.1.2",
Datacenter: "dc2",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.0.1.2", 8443,
structs.ServiceAddress{Address: "10.0.1.2", Port: 8443},
structs.ServiceAddress{Address: "198.18.1.2", Port: 443}),
},
}
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
func TestGatewayNodesDC3(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-1",
Node: "mesh-gateway",
Address: "10.30.1.1",
Datacenter: "dc3",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.30.1.1", 8443,
structs.ServiceAddress{Address: "10.0.1.1", Port: 8443},
structs.ServiceAddress{Address: "198.38.1.1", Port: 443}),
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "mesh-gateway-2",
Node: "mesh-gateway",
Address: "10.30.1.2",
Datacenter: "dc3",
},
Service: structs.TestNodeServiceMeshGatewayWithAddrs(t,
"10.30.1.2", 8443,
structs.ServiceAddress{Address: "10.30.1.2", Port: 8443},
structs.ServiceAddress{Address: "198.38.1.2", Port: 443}),
},
}
}
func TestGatewayServiceGroupBarDC1(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "bar-node-1",
Node: "bar-node-1",
Address: "10.1.1.4",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "bar-sidecar-proxy",
Address: "172.16.1.6",
Port: 2222,
Meta: map[string]string{
"version": "1",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "bar",
Upstreams: structs.TestUpstreams(t),
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "bar-node-2",
Node: "bar-node-2",
Address: "10.1.1.5",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "bar-sidecar-proxy",
Address: "172.16.1.7",
Port: 2222,
Meta: map[string]string{
"version": "1",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "bar",
Upstreams: structs.TestUpstreams(t),
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "bar-node-3",
Node: "bar-node-3",
Address: "10.1.1.6",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "bar-sidecar-proxy",
Address: "172.16.1.8",
Port: 2222,
Meta: map[string]string{
"version": "2",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "bar",
Upstreams: structs.TestUpstreams(t),
},
},
},
}
}
func TestGatewayServiceGroupFooDC1(t testing.T) structs.CheckServiceNodes {
return structs.CheckServiceNodes{
structs.CheckServiceNode{
Node: &structs.Node{
ID: "foo-node-1",
Node: "foo-node-1",
Address: "10.1.1.1",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "foo-sidecar-proxy",
Address: "172.16.1.3",
Port: 2222,
Meta: map[string]string{
"version": "1",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "foo",
Upstreams: structs.TestUpstreams(t),
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "foo-node-2",
Node: "foo-node-2",
Address: "10.1.1.2",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "foo-sidecar-proxy",
Address: "172.16.1.4",
Port: 2222,
Meta: map[string]string{
"version": "1",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "foo",
Upstreams: structs.TestUpstreams(t),
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "foo-node-3",
Node: "foo-node-3",
Address: "10.1.1.3",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "foo-sidecar-proxy",
Address: "172.16.1.5",
Port: 2222,
Meta: map[string]string{
"version": "2",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "foo",
Upstreams: structs.TestUpstreams(t),
},
},
},
structs.CheckServiceNode{
Node: &structs.Node{
ID: "foo-node-4",
Node: "foo-node-4",
Address: "10.1.1.7",
Datacenter: "dc1",
},
Service: &structs.NodeService{
Kind: structs.ServiceKindConnectProxy,
Service: "foo-sidecar-proxy",
Address: "172.16.1.9",
Port: 2222,
Meta: map[string]string{
"version": "2",
},
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "foo",
Upstreams: structs.TestUpstreams(t),
},
},
Checks: structs.HealthChecks{
&structs.HealthCheck{
Node: "foo-node-4",
ServiceName: "foo-sidecar-proxy",
Name: "proxy-alive",
Status: "warning",
},
},
},
}
}
// TestConfigSnapshot returns a fully populated snapshot
func TestConfigSnapshot(t testing.T) *ConfigSnapshot {
roots, leaf := TestCerts(t)
// no entries implies we'll get a default chain
dbChain := discoverychain.TestCompileConfigEntries(
t, "db", "default", "dc1",
connect.TestClusterID+".consul", "dc1", nil)
return &ConfigSnapshot{
Kind: structs.ServiceKindConnectProxy,
Service: "web-sidecar-proxy",
ProxyID: structs.NewServiceID("web-sidecar-proxy", nil),
Address: "0.0.0.0",
Port: 9999,
Proxy: structs.ConnectProxyConfig{
DestinationServiceID: "web",
DestinationServiceName: "web",
LocalServiceAddress: "127.0.0.1",
LocalServicePort: 8080,
Config: map[string]interface{}{
"foo": "bar",
},
Upstreams: structs.TestUpstreams(t),
},
Roots: roots,
ConnectProxy: configSnapshotConnectProxy{
ConfigSnapshotUpstreams: ConfigSnapshotUpstreams{
Leaf: leaf,
DiscoveryChain: map[string]*structs.CompiledDiscoveryChain{
"db": dbChain,
},
WatchedUpstreamEndpoints: map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"db.default.dc1": TestUpstreamNodes(t),
},
},
},
PreparedQueryEndpoints: map[string]structs.CheckServiceNodes{
"prepared_query:geo-cache": TestUpstreamNodes(t),
},
},
Datacenter: "dc1",
}
}
// TestConfigSnapshotDiscoveryChain returns a fully populated snapshot using a discovery chain
func TestConfigSnapshotDiscoveryChain(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "simple")
}
func TestConfigSnapshotDiscoveryChainExternalSNI(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "external-sni")
}
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
func TestConfigSnapshotDiscoveryChainWithOverrides(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "simple-with-overrides")
}
func TestConfigSnapshotDiscoveryChainWithFailover(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover")
}
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
func TestConfigSnapshotDiscoveryChainWithFailoverThroughRemoteGateway(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-remote-gateway")
}
func TestConfigSnapshotDiscoveryChainWithFailoverThroughRemoteGatewayTriggered(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-remote-gateway-triggered")
}
func TestConfigSnapshotDiscoveryChainWithDoubleFailoverThroughRemoteGateway(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-double-remote-gateway")
}
func TestConfigSnapshotDiscoveryChainWithDoubleFailoverThroughRemoteGatewayTriggered(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-double-remote-gateway-triggered")
}
func TestConfigSnapshotDiscoveryChainWithFailoverThroughLocalGateway(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-local-gateway")
}
func TestConfigSnapshotDiscoveryChainWithFailoverThroughLocalGatewayTriggered(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-local-gateway-triggered")
}
func TestConfigSnapshotDiscoveryChainWithDoubleFailoverThroughLocalGateway(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-double-local-gateway")
}
func TestConfigSnapshotDiscoveryChainWithDoubleFailoverThroughLocalGatewayTriggered(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "failover-through-double-local-gateway-triggered")
}
func TestConfigSnapshotDiscoveryChain_SplitterWithResolverRedirectMultiDC(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "splitter-with-resolver-redirect-multidc")
}
func TestConfigSnapshotDiscoveryChainWithEntries(t testing.T, additionalEntries ...structs.ConfigEntry) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "simple", additionalEntries...)
}
func TestConfigSnapshotDiscoveryChainDefault(t testing.T) *ConfigSnapshot {
return testConfigSnapshotDiscoveryChain(t, "default")
}
func testConfigSnapshotDiscoveryChain(t testing.T, variation string, additionalEntries ...structs.ConfigEntry) *ConfigSnapshot {
roots, leaf := TestCerts(t)
// Compile a chain.
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
var (
entries []structs.ConfigEntry
compileSetup func(req *discoverychain.CompileRequest)
)
switch variation {
case "default":
// no config entries
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
case "simple-with-overrides":
compileSetup = func(req *discoverychain.CompileRequest) {
req.OverrideMeshGateway.Mode = structs.MeshGatewayModeLocal
req.OverrideProtocol = "grpc"
req.OverrideConnectTimeout = 66 * time.Second
}
fallthrough
case "simple":
entries = append(entries,
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
},
)
case "external-sni":
entries = append(entries,
&structs.ServiceConfigEntry{
Kind: structs.ServiceDefaults,
Name: "db",
ExternalSNI: "db.some.other.service.mesh",
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
},
)
case "failover":
entries = append(entries,
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
Failover: map[string]structs.ServiceResolverFailover{
"*": {
Service: "fail",
},
},
},
)
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
case "failover-through-remote-gateway-triggered":
fallthrough
case "failover-through-remote-gateway":
entries = append(entries,
&structs.ServiceConfigEntry{
Kind: structs.ServiceDefaults,
Name: "db",
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeRemote,
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
Failover: map[string]structs.ServiceResolverFailover{
"*": {
Datacenters: []string{"dc2"},
},
},
},
)
case "failover-through-double-remote-gateway-triggered":
fallthrough
case "failover-through-double-remote-gateway":
entries = append(entries,
&structs.ServiceConfigEntry{
Kind: structs.ServiceDefaults,
Name: "db",
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeRemote,
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
Failover: map[string]structs.ServiceResolverFailover{
"*": {
Datacenters: []string{"dc2", "dc3"},
},
},
},
)
case "failover-through-local-gateway-triggered":
fallthrough
case "failover-through-local-gateway":
entries = append(entries,
&structs.ServiceConfigEntry{
Kind: structs.ServiceDefaults,
Name: "db",
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeLocal,
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
Failover: map[string]structs.ServiceResolverFailover{
"*": {
Datacenters: []string{"dc2"},
},
},
},
)
case "failover-through-double-local-gateway-triggered":
fallthrough
case "failover-through-double-local-gateway":
entries = append(entries,
&structs.ServiceConfigEntry{
Kind: structs.ServiceDefaults,
Name: "db",
MeshGateway: structs.MeshGatewayConfig{
Mode: structs.MeshGatewayModeLocal,
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
ConnectTimeout: 33 * time.Second,
Failover: map[string]structs.ServiceResolverFailover{
"*": {
Datacenters: []string{"dc2", "dc3"},
},
},
},
)
case "splitter-with-resolver-redirect-multidc":
entries = append(entries,
&structs.ProxyConfigEntry{
Kind: structs.ProxyDefaults,
Name: structs.ProxyConfigGlobal,
Config: map[string]interface{}{
"protocol": "http",
},
},
&structs.ServiceSplitterConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
Splits: []structs.ServiceSplit{
{Weight: 50, Service: "db-dc1"},
{Weight: 50, Service: "db-dc2"},
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db-dc1",
Redirect: &structs.ServiceResolverRedirect{
Service: "db",
ServiceSubset: "v1",
Datacenter: "dc1",
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db-dc2",
Redirect: &structs.ServiceResolverRedirect{
Service: "db",
ServiceSubset: "v2",
Datacenter: "dc2",
},
},
&structs.ServiceResolverConfigEntry{
Kind: structs.ServiceResolver,
Name: "db",
Subsets: map[string]structs.ServiceResolverSubset{
"v1": structs.ServiceResolverSubset{
Filter: "Service.Meta.version == v1",
},
"v2": structs.ServiceResolverSubset{
Filter: "Service.Meta.version == v2",
},
},
},
)
default:
t.Fatalf("unexpected variation: %q", variation)
return nil
}
if len(additionalEntries) > 0 {
entries = append(entries, additionalEntries...)
}
dbChain := discoverychain.TestCompileConfigEntries(t, "db", "default", "dc1", connect.TestClusterID+".consul", "dc1", compileSetup, entries...)
snap := &ConfigSnapshot{
Kind: structs.ServiceKindConnectProxy,
Service: "web-sidecar-proxy",
ProxyID: structs.NewServiceID("web-sidecar-proxy", nil),
Address: "0.0.0.0",
Port: 9999,
Proxy: structs.ConnectProxyConfig{
DestinationServiceID: "web",
DestinationServiceName: "web",
LocalServiceAddress: "127.0.0.1",
LocalServicePort: 8080,
Config: map[string]interface{}{
"foo": "bar",
},
Upstreams: structs.TestUpstreams(t),
},
Roots: roots,
ConnectProxy: configSnapshotConnectProxy{
ConfigSnapshotUpstreams: ConfigSnapshotUpstreams{
Leaf: leaf,
DiscoveryChain: map[string]*structs.CompiledDiscoveryChain{
"db": dbChain,
},
WatchedUpstreamEndpoints: map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"db.default.dc1": TestUpstreamNodes(t),
},
},
},
},
Datacenter: "dc1",
}
switch variation {
case "default":
connect: reconcile how upstream configuration works with discovery chains (#6225) * connect: reconcile how upstream configuration works with discovery chains The following upstream config fields for connect sidecars sanely integrate into discovery chain resolution: - Destination Namespace/Datacenter: Compilation occurs locally but using different default values for namespaces and datacenters. The xDS clusters that are created are named as they normally would be. - Mesh Gateway Mode (single upstream): If set this value overrides any value computed for any resolver for the entire discovery chain. The xDS clusters that are created may be named differently (see below). - Mesh Gateway Mode (whole sidecar): If set this value overrides any value computed for any resolver for the entire discovery chain. If this is specifically overridden for a single upstream this value is ignored in that case. The xDS clusters that are created may be named differently (see below). - Protocol (in opaque config): If set this value overrides the value computed when evaluating the entire discovery chain. If the normal chain would be TCP or if this override is set to TCP then the result is that we explicitly disable L7 Routing and Splitting. The xDS clusters that are created may be named differently (see below). - Connect Timeout (in opaque config): If set this value overrides the value for any resolver in the entire discovery chain. The xDS clusters that are created may be named differently (see below). If any of the above overrides affect the actual result of compiling the discovery chain (i.e. "tcp" becomes "grpc" instead of being a no-op override to "tcp") then the relevant parameters are hashed and provided to the xDS layer as a prefix for use in naming the Clusters. This is to ensure that if one Upstream discovery chain has no overrides and tangentially needs a cluster named "api.default.XXX", and another Upstream does have overrides for "api.default.XXX" that they won't cross-pollinate against the operator's wishes. Fixes #6159
2019-08-02 03:03:34 +00:00
case "simple-with-overrides":
case "simple":
case "external-sni":
case "failover":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["fail.default.dc1"] =
TestUpstreamNodesAlternate(t)
connect: fix failover through a mesh gateway to a remote datacenter (#6259) Failover is pushed entirely down to the data plane by creating envoy clusters and putting each successive destination in a different load assignment priority band. For example this shows that normally requests go to 1.2.3.4:8080 but when that fails they go to 6.7.8.9:8080: - name: foo load_assignment: cluster_name: foo policy: overprovisioning_factor: 100000 endpoints: - priority: 0 lb_endpoints: - endpoint: address: socket_address: address: 1.2.3.4 port_value: 8080 - priority: 1 lb_endpoints: - endpoint: address: socket_address: address: 6.7.8.9 port_value: 8080 Mesh gateways route requests based solely on the SNI header tacked onto the TLS layer. Envoy currently only lets you configure the outbound SNI header at the cluster layer. If you try to failover through a mesh gateway you ideally would configure the SNI value per endpoint, but that's not possible in envoy today. This PR introduces a simpler way around the problem for now: 1. We identify any target of failover that will use mesh gateway mode local or remote and then further isolate any resolver node in the compiled discovery chain that has a failover destination set to one of those targets. 2. For each of these resolvers we will perform a small measurement of comparative healths of the endpoints that come back from the health API for the set of primary target and serial failover targets. We walk the list of targets in order and if any endpoint is healthy we return that target, otherwise we move on to the next target. 3. The CDS and EDS endpoints both perform the measurements in (2) for the affected resolver nodes. 4. For CDS this measurement selects which TLS SNI field to use for the cluster (note the cluster is always going to be named for the primary target) 5. For EDS this measurement selects which set of endpoints will populate the cluster. Priority tiered failover is ignored. One of the big downsides to this approach to failover is that the failover detection and correction is going to be controlled by consul rather than deferring that entirely to the data plane as with the prior version. This also means that we are bound to only failover using official health signals and cannot make use of data plane signals like outlier detection to affect failover. In this specific scenario the lack of data plane signals is ok because the effectiveness is already muted by the fact that the ultimate destination endpoints will have their data plane signals scrambled when they pass through the mesh gateway wrapper anyway so we're not losing much. Another related fix is that we now use the endpoint health from the underlying service, not the health of the gateway (regardless of failover mode).
2019-08-05 18:30:35 +00:00
case "failover-through-remote-gateway-triggered":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc1"] =
TestUpstreamNodesInStatus(t, "critical")
fallthrough
case "failover-through-remote-gateway":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc2"] =
TestUpstreamNodesDC2(t)
snap.ConnectProxy.WatchedGatewayEndpoints = map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"dc2": TestGatewayNodesDC2(t),
},
}
case "failover-through-double-remote-gateway-triggered":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc1"] =
TestUpstreamNodesInStatus(t, "critical")
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc2"] =
TestUpstreamNodesInStatusDC2(t, "critical")
fallthrough
case "failover-through-double-remote-gateway":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc3"] = TestUpstreamNodesDC2(t)
snap.ConnectProxy.WatchedGatewayEndpoints = map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"dc2": TestGatewayNodesDC2(t),
"dc3": TestGatewayNodesDC3(t),
},
}
case "failover-through-local-gateway-triggered":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc1"] =
TestUpstreamNodesInStatus(t, "critical")
fallthrough
case "failover-through-local-gateway":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc2"] =
TestUpstreamNodesDC2(t)
snap.ConnectProxy.WatchedGatewayEndpoints = map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"dc1": TestGatewayNodesDC1(t),
},
}
case "failover-through-double-local-gateway-triggered":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc1"] =
TestUpstreamNodesInStatus(t, "critical")
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc2"] =
TestUpstreamNodesInStatusDC2(t, "critical")
fallthrough
case "failover-through-double-local-gateway":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"]["db.default.dc3"] = TestUpstreamNodesDC2(t)
snap.ConnectProxy.WatchedGatewayEndpoints = map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"dc1": TestGatewayNodesDC1(t),
},
}
case "splitter-with-resolver-redirect-multidc":
snap.ConnectProxy.WatchedUpstreamEndpoints["db"] = map[string]structs.CheckServiceNodes{
"v1.db.default.dc1": TestUpstreamNodes(t),
"v2.db.default.dc2": TestUpstreamNodesDC2(t),
}
default:
t.Fatalf("unexpected variation: %q", variation)
return nil
}
return snap
}
func TestConfigSnapshotMeshGateway(t testing.T) *ConfigSnapshot {
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
return testConfigSnapshotMeshGateway(t, true, false)
}
func TestConfigSnapshotMeshGatewayUsingFederationStates(t testing.T) *ConfigSnapshot {
return testConfigSnapshotMeshGateway(t, true, true)
}
func TestConfigSnapshotMeshGatewayNoServices(t testing.T) *ConfigSnapshot {
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
return testConfigSnapshotMeshGateway(t, false, false)
}
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
func testConfigSnapshotMeshGateway(t testing.T, populateServices bool, useFederationStates bool) *ConfigSnapshot {
roots, _ := TestCerts(t)
snap := &ConfigSnapshot{
Kind: structs.ServiceKindMeshGateway,
Service: "mesh-gateway",
ProxyID: structs.NewServiceID("mesh-gateway", nil),
Address: "1.2.3.4",
Port: 8443,
Proxy: structs.ConnectProxyConfig{
Config: map[string]interface{}{},
},
TaggedAddresses: map[string]structs.ServiceAddress{
structs.TaggedAddressLAN: structs.ServiceAddress{
Address: "1.2.3.4",
Port: 8443,
},
structs.TaggedAddressWAN: structs.ServiceAddress{
Address: "198.18.0.1",
Port: 443,
},
},
Roots: roots,
Datacenter: "dc1",
MeshGateway: configSnapshotMeshGateway{
WatchedServicesSet: true,
},
}
if populateServices {
snap.MeshGateway = configSnapshotMeshGateway{
WatchedServices: map[structs.ServiceID]context.CancelFunc{
structs.NewServiceID("foo", nil): nil,
structs.NewServiceID("bar", nil): nil,
},
WatchedServicesSet: true,
WatchedDatacenters: map[string]context.CancelFunc{
"dc2": nil,
},
ServiceGroups: map[structs.ServiceID]structs.CheckServiceNodes{
structs.NewServiceID("foo", nil): TestGatewayServiceGroupFooDC1(t),
structs.NewServiceID("bar", nil): TestGatewayServiceGroupBarDC1(t),
},
GatewayGroups: map[string]structs.CheckServiceNodes{
"dc2": TestGatewayNodesDC2(t),
},
}
wan federation via mesh gateways (#6884) This is like a Möbius strip of code due to the fact that low-level components (serf/memberlist) are connected to high-level components (the catalog and mesh-gateways) in a twisty maze of references which make it hard to dive into. With that in mind here's a high level summary of what you'll find in the patch: There are several distinct chunks of code that are affected: * new flags and config options for the server * retry join WAN is slightly different * retry join code is shared to discover primary mesh gateways from secondary datacenters * because retry join logic runs in the *agent* and the results of that operation for primary mesh gateways are needed in the *server* there are some methods like `RefreshPrimaryGatewayFallbackAddresses` that must occur at multiple layers of abstraction just to pass the data down to the right layer. * new cache type `FederationStateListMeshGatewaysName` for use in `proxycfg/xds` layers * the function signature for RPC dialing picked up a new required field (the node name of the destination) * several new RPCs for manipulating a FederationState object: `FederationState:{Apply,Get,List,ListMeshGateways}` * 3 read-only internal APIs for debugging use to invoke those RPCs from curl * raft and fsm changes to persist these FederationStates * replication for FederationStates as they are canonically stored in the Primary and replicated to the Secondaries. * a special derivative of anti-entropy that runs in secondaries to snapshot their local mesh gateway `CheckServiceNodes` and sync them into their upstream FederationState in the primary (this works in conjunction with the replication to distribute addresses for all mesh gateways in all DCs to all other DCs) * a "gateway locator" convenience object to make use of this data to choose the addresses of gateways to use for any given RPC or gossip operation to a remote DC. This gets data from the "retry join" logic in the agent and also directly calls into the FSM. * RPC (`:8300`) on the server sniffs the first byte of a new connection to determine if it's actually doing native TLS. If so it checks the ALPN header for protocol determination (just like how the existing system uses the type-byte marker). * 2 new kinds of protocols are exclusively decoded via this native TLS mechanism: one for ferrying "packet" operations (udp-like) from the gossip layer and one for "stream" operations (tcp-like). The packet operations re-use sockets (using length-prefixing) to cut down on TLS re-negotiation overhead. * the server instances specially wrap the `memberlist.NetTransport` when running with gateway federation enabled (in a `wanfed.Transport`). The general gist is that if it tries to dial a node in the SAME datacenter (deduced by looking at the suffix of the node name) there is no change. If dialing a DIFFERENT datacenter it is wrapped up in a TLS+ALPN blob and sent through some mesh gateways to eventually end up in a server's :8300 port. * a new flag when launching a mesh gateway via `consul connect envoy` to indicate that the servers are to be exposed. This sets a special service meta when registering the gateway into the catalog. * `proxycfg/xds` notice this metadata blob to activate additional watches for the FederationState objects as well as the location of all of the consul servers in that datacenter. * `xds:` if the extra metadata is in place additional clusters are defined in a DC to bulk sink all traffic to another DC's gateways. For the current datacenter we listen on a wildcard name (`server.<dc>.consul`) that load balances all servers as well as one mini-cluster per node (`<node>.server.<dc>.consul`) * the `consul tls cert create` command got a new flag (`-node`) to help create an additional SAN in certs that can be used with this flavor of federation.
2020-03-09 20:59:02 +00:00
if useFederationStates {
snap.MeshGateway.FedStateGateways = map[string]structs.CheckServiceNodes{
"dc2": TestGatewayNodesDC2(t),
}
delete(snap.MeshGateway.GatewayGroups, "dc2")
}
}
return snap
}
func TestConfigSnapshotIngressGateway(t testing.T) *ConfigSnapshot {
return testConfigSnapshotIngressGateway(t, true)
}
func TestConfigSnapshotIngressGatewayNoServices(t testing.T) *ConfigSnapshot {
return testConfigSnapshotIngressGateway(t, false)
}
func testConfigSnapshotIngressGateway(t testing.T, populateServices bool) *ConfigSnapshot {
roots, leaf := TestCerts(t)
dbChain := discoverychain.TestCompileConfigEntries(
t, "db", "default", "dc1",
connect.TestClusterID+".consul", "dc1", nil)
snap := &ConfigSnapshot{
Kind: structs.ServiceKindIngressGateway,
Service: "ingress-gateway",
ProxyID: structs.NewServiceID("ingress-gateway", nil),
Address: "1.2.3.4",
Roots: roots,
Datacenter: "dc1",
}
if populateServices {
snap.IngressGateway = configSnapshotIngressGateway{
ConfigSnapshotUpstreams: ConfigSnapshotUpstreams{
Leaf: leaf,
DiscoveryChain: map[string]*structs.CompiledDiscoveryChain{
"db": dbChain,
},
WatchedUpstreamEndpoints: map[string]map[string]structs.CheckServiceNodes{
"db": map[string]structs.CheckServiceNodes{
"db.default.dc1": TestUpstreamNodes(t),
},
},
},
Upstreams: structs.Upstreams{
{
// We rely on this one having default type in a few tests...
DestinationName: "db",
LocalBindPort: 9191,
LocalBindAddress: "2.3.4.5",
},
},
}
}
return snap
}
func TestConfigSnapshotExposeConfig(t testing.T) *ConfigSnapshot {
return &ConfigSnapshot{
Kind: structs.ServiceKindConnectProxy,
Service: "web-proxy",
ProxyID: structs.NewServiceID("web-proxy", nil),
Address: "1.2.3.4",
Port: 8080,
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "web",
DestinationServiceID: "web",
LocalServicePort: 8080,
Expose: structs.ExposeConfig{
Checks: false,
Paths: []structs.ExposePath{
{
LocalPathPort: 8080,
Path: "/health1",
ListenerPort: 21500,
},
{
LocalPathPort: 8080,
Path: "/health2",
ListenerPort: 21501,
},
},
},
},
Datacenter: "dc1",
}
}
func TestConfigSnapshotGRPCExposeHTTP1(t testing.T) *ConfigSnapshot {
return &ConfigSnapshot{
Kind: structs.ServiceKindConnectProxy,
Service: "grpc-proxy",
ProxyID: structs.NewServiceID("grpc-proxy", nil),
Address: "1.2.3.4",
Port: 8080,
Proxy: structs.ConnectProxyConfig{
DestinationServiceName: "grpc",
DestinationServiceID: "grpc",
LocalServicePort: 8080,
Config: map[string]interface{}{
"protocol": "grpc",
},
Expose: structs.ExposeConfig{
Checks: false,
Paths: []structs.ExposePath{
{
LocalPathPort: 8090,
Path: "/healthz",
ListenerPort: 21500,
Protocol: "http",
},
},
},
},
Datacenter: "dc1",
}
}
// ControllableCacheType is a cache.Type that simulates a typical blocking RPC
// but lets us control the responses and when they are delivered easily.
type ControllableCacheType struct {
index uint64
value sync.Map
// Need a condvar to trigger all blocking requests (there might be multiple
// for same type due to background refresh and timing issues) when values
// change. Chans make it nondeterministic which one triggers or need extra
// locking to coordinate replacing after close etc.
triggerMu sync.Mutex
trigger *sync.Cond
blocking bool
lastReq atomic.Value
}
// NewControllableCacheType returns a cache.Type that can be controlled for
// testing.
func NewControllableCacheType(t testing.T) *ControllableCacheType {
c := &ControllableCacheType{
index: 5,
blocking: true,
}
c.trigger = sync.NewCond(&c.triggerMu)
return c
}
// Set sets the response value to be returned from subsequent cache gets for the
// type.
func (ct *ControllableCacheType) Set(key string, value interface{}) {
atomic.AddUint64(&ct.index, 1)
ct.value.Store(key, value)
ct.triggerMu.Lock()
ct.trigger.Broadcast()
ct.triggerMu.Unlock()
}
// Fetch implements cache.Type. It simulates blocking or non-blocking queries.
func (ct *ControllableCacheType) Fetch(opts cache.FetchOptions, req cache.Request) (cache.FetchResult, error) {
index := atomic.LoadUint64(&ct.index)
ct.lastReq.Store(req)
shouldBlock := ct.blocking && opts.MinIndex > 0 && opts.MinIndex == index
if shouldBlock {
// Wait for return to be triggered. We ignore timeouts based on opts.Timeout
// since in practice they will always be way longer than our tests run for
// and the caller can simulate timeout by triggering return without changing
// index or value.
ct.triggerMu.Lock()
ct.trigger.Wait()
ct.triggerMu.Unlock()
}
info := req.CacheInfo()
key := path.Join(info.Key, info.Datacenter) // omit token for testing purposes
// reload index as it probably got bumped
index = atomic.LoadUint64(&ct.index)
val, _ := ct.value.Load(key)
if err, ok := val.(error); ok {
return cache.FetchResult{
Value: nil,
Index: index,
}, err
}
return cache.FetchResult{
Value: val,
Index: index,
}, nil
}
func (ct *ControllableCacheType) RegisterOptions() cache.RegisterOptions {
return cache.RegisterOptions{
Refresh: ct.blocking,
SupportsBlocking: ct.blocking,
RefreshTimeout: 10 * time.Minute,
}
}