mirror of https://github.com/status-im/consul.git
219 lines
6.0 KiB
Go
219 lines
6.0 KiB
Go
// Copyright (c) HashiCorp, Inc.
|
|
// SPDX-License-Identifier: BUSL-1.1
|
|
|
|
package peering
|
|
|
|
import (
|
|
"fmt"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/mitchellh/copystructure"
|
|
"github.com/stretchr/testify/assert"
|
|
"github.com/stretchr/testify/require"
|
|
|
|
"github.com/hashicorp/consul/api"
|
|
"github.com/hashicorp/consul/sdk/testutil/retry"
|
|
"github.com/hashicorp/consul/test/integration/consul-container/libs/utils"
|
|
"github.com/hashicorp/consul/testing/deployer/topology"
|
|
)
|
|
|
|
// TestAC7_2RotateLeader ensures that after a leader rotation, information continues to replicate to peers
|
|
// NOTE: because suiteRotateLeader needs to mutate the topo, we actually *DO NOT* share a topo
|
|
type ac7_2RotateLeaderSuite struct {
|
|
DC string
|
|
Peer string
|
|
|
|
sidServer topology.ID
|
|
nodeServer topology.NodeID
|
|
|
|
sidClient topology.ID
|
|
nodeClient topology.NodeID
|
|
|
|
upstream *topology.Upstream
|
|
}
|
|
|
|
func TestAC7_2RotateLeader(t *testing.T) {
|
|
suites := []*ac7_2RotateLeaderSuite{
|
|
{DC: "dc1", Peer: "dc2"},
|
|
{DC: "dc2", Peer: "dc1"},
|
|
}
|
|
ct := NewCommonTopo(t)
|
|
for _, s := range suites {
|
|
s.setup(t, ct)
|
|
}
|
|
ct.Launch(t)
|
|
for _, s := range suites {
|
|
s := s
|
|
t.Run(fmt.Sprintf("%s->%s", s.DC, s.Peer), func(t *testing.T) {
|
|
// no t.Parallel() due to Relaunch
|
|
s.test(t, ct)
|
|
})
|
|
}
|
|
}
|
|
|
|
// makes client in clu, server in peerClu
|
|
func (s *ac7_2RotateLeaderSuite) setup(t *testing.T, ct *commonTopo) {
|
|
const prefix = "ac7-2-"
|
|
|
|
clu := ct.ClusterByDatacenter(t, s.DC)
|
|
peerClu := ct.ClusterByDatacenter(t, s.Peer)
|
|
partition := "default"
|
|
peer := LocalPeerName(peerClu, "default")
|
|
cluPeerName := LocalPeerName(clu, "default")
|
|
|
|
server := NewFortioServiceWithDefaults(
|
|
peerClu.Datacenter,
|
|
topology.ID{
|
|
Name: prefix + "server-http",
|
|
Partition: partition,
|
|
},
|
|
nil,
|
|
)
|
|
|
|
// Make clients which have server upstreams
|
|
upstream := &topology.Upstream{
|
|
ID: topology.ID{
|
|
Name: server.ID.Name,
|
|
Partition: partition,
|
|
},
|
|
LocalPort: 5001,
|
|
Peer: peer,
|
|
}
|
|
// create client in us
|
|
client := NewFortioServiceWithDefaults(
|
|
clu.Datacenter,
|
|
topology.ID{
|
|
Name: prefix + "client",
|
|
Partition: partition,
|
|
},
|
|
func(s *topology.Workload) {
|
|
s.Upstreams = []*topology.Upstream{
|
|
upstream,
|
|
}
|
|
},
|
|
)
|
|
clientNode := ct.AddServiceNode(clu, serviceExt{Workload: client,
|
|
Config: &api.ServiceConfigEntry{
|
|
Kind: api.ServiceDefaults,
|
|
Name: client.ID.Name,
|
|
Partition: ConfigEntryPartition(client.ID.Partition),
|
|
Protocol: "http",
|
|
UpstreamConfig: &api.UpstreamConfiguration{
|
|
Defaults: &api.UpstreamConfig{
|
|
MeshGateway: api.MeshGatewayConfig{
|
|
Mode: api.MeshGatewayModeLocal,
|
|
},
|
|
},
|
|
},
|
|
},
|
|
})
|
|
// actually to be used by the other pairing
|
|
serverNode := ct.AddServiceNode(peerClu, serviceExt{
|
|
Workload: server,
|
|
Config: &api.ServiceConfigEntry{
|
|
Kind: api.ServiceDefaults,
|
|
Name: server.ID.Name,
|
|
Partition: ConfigEntryPartition(partition),
|
|
Protocol: "http",
|
|
},
|
|
Exports: []api.ServiceConsumer{{Peer: cluPeerName}},
|
|
Intentions: &api.ServiceIntentionsConfigEntry{
|
|
Kind: api.ServiceIntentions,
|
|
Name: server.ID.Name,
|
|
Partition: ConfigEntryPartition(partition),
|
|
Sources: []*api.SourceIntention{
|
|
{
|
|
Name: client.ID.Name,
|
|
Peer: cluPeerName,
|
|
Action: api.IntentionActionAllow,
|
|
},
|
|
},
|
|
},
|
|
})
|
|
|
|
s.sidClient = client.ID
|
|
s.nodeClient = clientNode.ID()
|
|
s.upstream = upstream
|
|
s.sidServer = server.ID
|
|
s.nodeServer = serverNode.ID()
|
|
}
|
|
|
|
func (s *ac7_2RotateLeaderSuite) test(t *testing.T, ct *commonTopo) {
|
|
dc := ct.Sprawl.Topology().Clusters[s.DC]
|
|
peer := ct.Sprawl.Topology().Clusters[s.Peer]
|
|
clDC := ct.APIClientForCluster(t, dc)
|
|
clPeer := ct.APIClientForCluster(t, peer)
|
|
|
|
svcServer := peer.WorkloadByID(s.nodeServer, s.sidServer)
|
|
svcClient := dc.WorkloadByID(s.nodeClient, s.sidClient)
|
|
ct.Assert.HealthyWithPeer(t, dc.Name, svcServer.ID, LocalPeerName(peer, "default"))
|
|
|
|
ct.Assert.FortioFetch2HeaderEcho(t, svcClient, s.upstream)
|
|
|
|
// force leader election
|
|
rotateLeader(t, clDC)
|
|
rotateLeader(t, clPeer)
|
|
|
|
// unexport httpServer
|
|
ce, _, err := clPeer.ConfigEntries().Get(api.ExportedServices, s.sidServer.Partition, nil)
|
|
require.NoError(t, err)
|
|
// ceAsES = config entry as ExportedServicesConfigEntry
|
|
ceAsES := ce.(*api.ExportedServicesConfigEntry)
|
|
origCE, err := copystructure.Copy(ceAsES)
|
|
require.NoError(t, err)
|
|
found := 0
|
|
foundI := 0
|
|
for i, svc := range ceAsES.Services {
|
|
if svc.Name == s.sidServer.Name && utils.DefaultToEmpty(svc.Namespace) == utils.DefaultToEmpty(s.sidServer.Namespace) {
|
|
found += 1
|
|
foundI = i
|
|
}
|
|
}
|
|
require.Equal(t, found, 1)
|
|
// remove found entry
|
|
ceAsES.Services = append(ceAsES.Services[:foundI], ceAsES.Services[foundI+1:]...)
|
|
_, _, err = clPeer.ConfigEntries().Set(ceAsES, nil)
|
|
require.NoError(t, err)
|
|
t.Cleanup(func() {
|
|
// restore for next pairing
|
|
_, _, err = clPeer.ConfigEntries().Set(origCE.(*api.ExportedServicesConfigEntry), nil)
|
|
require.NoError(t, err)
|
|
})
|
|
|
|
// expect health entry in for peer to disappear
|
|
retry.RunWith(&retry.Timer{Timeout: time.Minute, Wait: time.Millisecond * 500}, t, func(r *retry.R) {
|
|
svcs, _, err := clDC.Health().Service(s.sidServer.Name, "", true, utils.CompatQueryOpts(&api.QueryOptions{
|
|
Partition: s.sidServer.Partition,
|
|
Namespace: s.sidServer.Namespace,
|
|
Peer: LocalPeerName(peer, "default"),
|
|
}))
|
|
require.NoError(r, err)
|
|
assert.Equal(r, len(svcs), 0, "health entry for imported service gone")
|
|
})
|
|
}
|
|
|
|
func rotateLeader(t *testing.T, cl *api.Client) {
|
|
t.Helper()
|
|
oldLeader := findLeader(t, cl)
|
|
_, err := cl.Operator().RaftLeaderTransfer("", nil)
|
|
require.NoError(t, err)
|
|
retry.RunWith(&retry.Timer{Timeout: 30 * time.Second, Wait: time.Second}, t, func(r *retry.R) {
|
|
newLeader := findLeader(r, cl)
|
|
require.NotEqual(r, oldLeader.ID, newLeader.ID)
|
|
})
|
|
}
|
|
|
|
func findLeader(t require.TestingT, cl *api.Client) *api.RaftServer {
|
|
raftConfig, err := cl.Operator().RaftGetConfiguration(nil)
|
|
require.NoError(t, err)
|
|
var leader *api.RaftServer
|
|
for _, svr := range raftConfig.Servers {
|
|
if svr.Leader {
|
|
leader = svr
|
|
}
|
|
}
|
|
require.NotNil(t, leader)
|
|
return leader
|
|
}
|