ef0999547a
Add a skip condition to all tests slower than 100ms. This change was made using `gotestsum tool slowest` with data from the last 3 CI runs of master. See https://github.com/gotestyourself/gotestsum#finding-and-skipping-slow-tests With this change: ``` $ time go test -count=1 -short ./agent ok github.com/hashicorp/consul/agent 0.743s real 0m4.791s $ time go test -count=1 -short ./agent/consul ok github.com/hashicorp/consul/agent/consul 4.229s real 0m8.769s ```
525 lines
12 KiB
Go
525 lines
12 KiB
Go
package consul
|
|
|
|
import (
|
|
"context"
|
|
"os"
|
|
"testing"
|
|
"time"
|
|
|
|
"github.com/hashicorp/consul/agent/structs"
|
|
"github.com/hashicorp/consul/sdk/testutil/retry"
|
|
"github.com/hashicorp/consul/testrpc"
|
|
"github.com/hashicorp/raft"
|
|
"github.com/hashicorp/serf/serf"
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
func TestAutopilot_IdempotentShutdown(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
dir1, s1 := testServerWithConfig(t, nil)
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
retry.Run(t, func(r *retry.R) { r.Check(waitForLeader(s1)) })
|
|
|
|
s1.autopilot.Start(context.Background())
|
|
s1.autopilot.Start(context.Background())
|
|
s1.autopilot.Start(context.Background())
|
|
<-s1.autopilot.Stop()
|
|
<-s1.autopilot.Stop()
|
|
<-s1.autopilot.Stop()
|
|
}
|
|
|
|
func TestAutopilot_CleanupDeadServer(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
dc := "dc1"
|
|
conf := func(c *Config) {
|
|
c.Datacenter = dc
|
|
c.Bootstrap = false
|
|
c.BootstrapExpect = 5
|
|
}
|
|
dir1, s1 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
dir2, s2 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
dir3, s3 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir3)
|
|
defer s3.Shutdown()
|
|
|
|
dir4, s4 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir4)
|
|
defer s4.Shutdown()
|
|
|
|
dir5, s5 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir5)
|
|
defer s5.Shutdown()
|
|
|
|
servers := []*Server{s1, s2, s3, s4, s5}
|
|
|
|
// Try to join
|
|
joinLAN(t, s2, s1)
|
|
joinLAN(t, s3, s1)
|
|
joinLAN(t, s4, s1)
|
|
joinLAN(t, s5, s1)
|
|
|
|
for _, s := range servers {
|
|
testrpc.WaitForLeader(t, s.RPC, dc)
|
|
retry.Run(t, func(r *retry.R) { r.Check(wantPeers(s, 5)) })
|
|
}
|
|
|
|
require := require.New(t)
|
|
testrpc.WaitForLeader(t, s1.RPC, "dc1")
|
|
leaderIndex := -1
|
|
for i, s := range servers {
|
|
if s.IsLeader() {
|
|
leaderIndex = i
|
|
break
|
|
}
|
|
}
|
|
require.NotEqual(leaderIndex, -1)
|
|
|
|
// Shutdown two non-leader servers
|
|
killed := make(map[string]struct{})
|
|
for i, s := range servers {
|
|
if i != leaderIndex {
|
|
s.Shutdown()
|
|
killed[string(s.config.NodeID)] = struct{}{}
|
|
}
|
|
if len(killed) == 2 {
|
|
break
|
|
}
|
|
}
|
|
|
|
retry.Run(t, func(r *retry.R) {
|
|
alive := 0
|
|
for _, m := range servers[leaderIndex].LANMembers() {
|
|
if m.Status == serf.StatusAlive {
|
|
alive++
|
|
}
|
|
}
|
|
if alive != 3 {
|
|
r.Fatalf("Expected three alive servers instead of %d", alive)
|
|
}
|
|
})
|
|
|
|
// Make sure the dead servers are removed and we're back to 3 total peers
|
|
for _, s := range servers {
|
|
_, killed := killed[string(s.config.NodeID)]
|
|
if !killed {
|
|
retry.Run(t, func(r *retry.R) { r.Check(wantPeers(s, 3)) })
|
|
}
|
|
}
|
|
}
|
|
|
|
func TestAutopilot_CleanupDeadNonvoter(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
dir1, s1 := testServerWithConfig(t, func(c *Config) {
|
|
c.AutopilotConfig = &structs.AutopilotConfig{
|
|
CleanupDeadServers: true,
|
|
ServerStabilizationTime: 100 * time.Millisecond,
|
|
}
|
|
})
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
// we have to wait for autopilot to be running long enough for the server stabilization time
|
|
// to kick in for this test to work.
|
|
time.Sleep(100 * time.Millisecond)
|
|
|
|
dir2, s2 := testServerDCBootstrap(t, "dc1", false)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
testrpc.WaitForLeader(t, s1.RPC, "dc1")
|
|
|
|
// Have s2 join and then shut it down immediately before it gets a chance to
|
|
// be promoted to a voter.
|
|
joinLAN(t, s2, s1)
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft([]*Server{s1, s2}))
|
|
})
|
|
s2.Shutdown()
|
|
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft([]*Server{s1}))
|
|
})
|
|
}
|
|
|
|
func TestAutopilot_CleanupDeadServerPeriodic(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
t.Parallel()
|
|
dir1, s1 := testServerWithConfig(t, func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = true
|
|
})
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
conf := func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = false
|
|
}
|
|
|
|
dir2, s2 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
dir3, s3 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir3)
|
|
defer s3.Shutdown()
|
|
|
|
dir4, s4 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir4)
|
|
defer s4.Shutdown()
|
|
|
|
dir5, s5 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir5)
|
|
defer s5.Shutdown()
|
|
|
|
servers := []*Server{s1, s2, s3, s4, s5}
|
|
|
|
// Join the servers to s1, and wait until they are all promoted to
|
|
// voters.
|
|
for _, s := range servers[1:] {
|
|
joinLAN(t, s, s1)
|
|
}
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft(servers))
|
|
for _, s := range servers {
|
|
r.Check(wantPeers(s, 5))
|
|
}
|
|
})
|
|
|
|
// Kill a non-leader server
|
|
s4.Shutdown()
|
|
|
|
// Should be removed from the peers automatically
|
|
servers = []*Server{s1, s2, s3, s5}
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft(servers))
|
|
for _, s := range servers {
|
|
r.Check(wantPeers(s, 4))
|
|
}
|
|
})
|
|
}
|
|
|
|
func TestAutopilot_RollingUpdate(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
t.Parallel()
|
|
dir1, s1 := testServerWithConfig(t, func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = true
|
|
})
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
conf := func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = false
|
|
}
|
|
|
|
dir2, s2 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
dir3, s3 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir3)
|
|
defer s3.Shutdown()
|
|
|
|
// Join the servers to s1, and wait until they are all promoted to
|
|
// voters.
|
|
servers := []*Server{s1, s2, s3}
|
|
for _, s := range servers[1:] {
|
|
joinLAN(t, s, s1)
|
|
}
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft(servers))
|
|
for _, s := range servers {
|
|
r.Check(wantPeers(s, 3))
|
|
}
|
|
})
|
|
|
|
// Add one more server like we are doing a rolling update.
|
|
dir4, s4 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir4)
|
|
defer s4.Shutdown()
|
|
joinLAN(t, s1, s4)
|
|
servers = append(servers, s4)
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft(servers))
|
|
for _, s := range servers {
|
|
r.Check(wantPeers(s, 3))
|
|
}
|
|
})
|
|
|
|
// Now kill one of the "old" nodes like we are doing a rolling update.
|
|
s3.Shutdown()
|
|
|
|
isVoter := func() bool {
|
|
future := s1.raft.GetConfiguration()
|
|
if err := future.Error(); err != nil {
|
|
t.Fatalf("err: %v", err)
|
|
}
|
|
for _, s := range future.Configuration().Servers {
|
|
if string(s.ID) == string(s4.config.NodeID) {
|
|
return s.Suffrage == raft.Voter
|
|
}
|
|
}
|
|
t.Fatalf("didn't find s4")
|
|
return false
|
|
}
|
|
|
|
// Wait for s4 to stabilize, get promoted to a voter, and for s3 to be
|
|
// removed.
|
|
servers = []*Server{s1, s2, s4}
|
|
retry.Run(t, func(r *retry.R) {
|
|
r.Check(wantRaft(servers))
|
|
for _, s := range servers {
|
|
r.Check(wantPeers(s, 3))
|
|
}
|
|
if !isVoter() {
|
|
r.Fatalf("should be a voter")
|
|
}
|
|
})
|
|
}
|
|
|
|
func TestAutopilot_CleanupStaleRaftServer(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
t.Parallel()
|
|
dir1, s1 := testServerDCBootstrap(t, "dc1", true)
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
dir2, s2 := testServerDCBootstrap(t, "dc1", false)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
dir3, s3 := testServerDCBootstrap(t, "dc1", false)
|
|
defer os.RemoveAll(dir3)
|
|
defer s3.Shutdown()
|
|
|
|
dir4, s4 := testServerDCBootstrap(t, "dc1", false)
|
|
defer os.RemoveAll(dir4)
|
|
defer s4.Shutdown()
|
|
|
|
servers := []*Server{s1, s2, s3}
|
|
|
|
// Join the servers to s1
|
|
for _, s := range servers[1:] {
|
|
joinLAN(t, s, s1)
|
|
}
|
|
|
|
for _, s := range servers {
|
|
retry.Run(t, func(r *retry.R) { r.Check(wantPeers(s, 3)) })
|
|
}
|
|
|
|
testrpc.WaitForLeader(t, s1.RPC, "dc1")
|
|
|
|
// Add s4 to peers directly
|
|
addVoterFuture := s1.raft.AddVoter(raft.ServerID(s4.config.NodeID), raft.ServerAddress(joinAddrLAN(s4)), 0, 0)
|
|
if err := addVoterFuture.Error(); err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
|
|
// Verify we have 4 peers
|
|
peers, err := s1.autopilot.NumVoters()
|
|
if err != nil {
|
|
t.Fatal(err)
|
|
}
|
|
if peers != 4 {
|
|
t.Fatalf("bad: %v", peers)
|
|
}
|
|
|
|
// Wait for s4 to be removed
|
|
for _, s := range []*Server{s1, s2, s3} {
|
|
retry.Run(t, func(r *retry.R) { r.Check(wantPeers(s, 3)) })
|
|
}
|
|
}
|
|
|
|
func TestAutopilot_PromoteNonVoter(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
t.Parallel()
|
|
dir1, s1 := testServerWithConfig(t, func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = true
|
|
c.AutopilotConfig.ServerStabilizationTime = 200 * time.Millisecond
|
|
c.ServerHealthInterval = 100 * time.Millisecond
|
|
c.AutopilotInterval = 100 * time.Millisecond
|
|
})
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
codec := rpcClient(t, s1)
|
|
defer codec.Close()
|
|
testrpc.WaitForLeader(t, s1.RPC, "dc1")
|
|
|
|
// this may seem arbitrary but we need to get past the server stabilization time
|
|
// so that we start factoring in that time for newly connected nodes.
|
|
time.Sleep(100 * time.Millisecond)
|
|
|
|
dir2, s2 := testServerWithConfig(t, func(c *Config) {
|
|
c.Datacenter = "dc1"
|
|
c.Bootstrap = false
|
|
c.RaftConfig.ProtocolVersion = 3
|
|
})
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
joinLAN(t, s2, s1)
|
|
|
|
// Make sure we see it as a nonvoter initially. We wait until half
|
|
// the stabilization period has passed.
|
|
retry.Run(t, func(r *retry.R) {
|
|
future := s1.raft.GetConfiguration()
|
|
if err := future.Error(); err != nil {
|
|
r.Fatal(err)
|
|
}
|
|
|
|
servers := future.Configuration().Servers
|
|
if len(servers) != 2 {
|
|
r.Fatalf("bad: %v", servers)
|
|
}
|
|
if servers[1].Suffrage != raft.Nonvoter {
|
|
r.Fatalf("bad: %v", servers)
|
|
}
|
|
health := s1.autopilot.GetServerHealth(servers[1].ID)
|
|
if health == nil {
|
|
r.Fatal("nil health")
|
|
}
|
|
if !health.Healthy {
|
|
r.Fatalf("bad: %v", health)
|
|
}
|
|
if time.Since(health.StableSince) < s1.config.AutopilotConfig.ServerStabilizationTime/2 {
|
|
r.Fatal("stable period not elapsed")
|
|
}
|
|
})
|
|
|
|
// Make sure it ends up as a voter.
|
|
retry.Run(t, func(r *retry.R) {
|
|
future := s1.raft.GetConfiguration()
|
|
if err := future.Error(); err != nil {
|
|
r.Fatal(err)
|
|
}
|
|
|
|
servers := future.Configuration().Servers
|
|
if len(servers) != 2 {
|
|
r.Fatalf("bad: %v", servers)
|
|
}
|
|
if servers[1].Suffrage != raft.Voter {
|
|
r.Fatalf("bad: %v", servers)
|
|
}
|
|
})
|
|
}
|
|
|
|
func TestAutopilot_MinQuorum(t *testing.T) {
|
|
if testing.Short() {
|
|
t.Skip("too slow for testing.Short")
|
|
}
|
|
|
|
dc := "dc1"
|
|
conf := func(c *Config) {
|
|
c.Datacenter = dc
|
|
c.Bootstrap = false
|
|
c.BootstrapExpect = 4
|
|
c.AutopilotConfig.MinQuorum = 3
|
|
c.AutopilotInterval = 100 * time.Millisecond
|
|
}
|
|
dir1, s1 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir1)
|
|
defer s1.Shutdown()
|
|
|
|
dir2, s2 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir2)
|
|
defer s2.Shutdown()
|
|
|
|
dir3, s3 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir3)
|
|
defer s3.Shutdown()
|
|
|
|
dir4, s4 := testServerWithConfig(t, conf)
|
|
defer os.RemoveAll(dir4)
|
|
defer s4.Shutdown()
|
|
|
|
servers := map[string]*Server{s1.config.NodeName: s1,
|
|
s2.config.NodeName: s2,
|
|
s3.config.NodeName: s3,
|
|
s4.config.NodeName: s4}
|
|
|
|
// Try to join
|
|
joinLAN(t, s2, s1)
|
|
joinLAN(t, s3, s1)
|
|
joinLAN(t, s4, s1)
|
|
|
|
//Differentiate between leader and server
|
|
findStatus := func(leader bool) *Server {
|
|
for _, mem := range servers {
|
|
if mem.IsLeader() == leader {
|
|
return mem
|
|
}
|
|
if !mem.IsLeader() == !leader {
|
|
return mem
|
|
}
|
|
}
|
|
|
|
return nil
|
|
}
|
|
testrpc.WaitForLeader(t, s1.RPC, dc)
|
|
|
|
// Have autopilot take one into left
|
|
dead := findStatus(false)
|
|
if dead == nil {
|
|
t.Fatalf("no members set")
|
|
}
|
|
require.NoError(t, dead.Shutdown())
|
|
retry.Run(t, func(r *retry.R) {
|
|
leader := findStatus(true)
|
|
if leader == nil {
|
|
r.Fatalf("no members set")
|
|
}
|
|
for _, m := range leader.LANMembers() {
|
|
if m.Name == dead.config.NodeName && m.Status != serf.StatusLeft {
|
|
r.Fatalf("%v should be left, got %v", m.Name, m.Status.String())
|
|
}
|
|
}
|
|
})
|
|
|
|
delete(servers, dead.config.NodeName)
|
|
//Autopilot should not take this one into left
|
|
dead = findStatus(false)
|
|
require.NoError(t, dead.Shutdown())
|
|
|
|
retry.Run(t, func(r *retry.R) {
|
|
leader := findStatus(true)
|
|
if leader == nil {
|
|
r.Fatalf("no members set")
|
|
}
|
|
for _, m := range leader.LANMembers() {
|
|
if m.Name == dead.config.NodeName && m.Status != serf.StatusFailed {
|
|
r.Fatalf("%v should be failed, got %v", m.Name, m.Status.String())
|
|
}
|
|
}
|
|
})
|
|
}
|