func main() {
defer glog.Flush()
prg := &program{}
if err := svc.Run(prg, os.Interrupt, syscall.SIGTERM, syscall.SIGHUP, syscall.SIGQUIT, syscall.SIGINT); err != nil {
log.Fatal(err)
}
}
func main() {
glog.InitWithFlag(flagSet)
flagSet.Parse(os.Args[1:])
if *showVersion {
fmt.Println(version.String("nsqadmin"))
return
}
defer glog.Flush()
if *templateDir != "" {
log.Printf("WARNING: --template-dir is deprecated and will be removed in the next release (templates are now compiled into the binary)")
}
exitChan := make(chan int)
signalChan := make(chan os.Signal, 1)
go func() {
<-signalChan
exitChan <- 1
}()
signal.Notify(signalChan, syscall.SIGINT, syscall.SIGTERM)
var cfg map[string]interface{}
if *config != "" {
_, err := toml.DecodeFile(*config, &cfg)
if err != nil {
log.Fatalf("ERROR: failed to load config file %s - %s", *config, err)
}
}
opts := nsqadmin.NewOptions()
options.Resolve(opts, flagSet, cfg)
if opts.LogDir != "" {
glog.SetGLogDir(opts.LogDir)
}
glog.StartWorker(time.Second * 2)
nsqadmin := nsqadmin.New(opts)
nsqadmin.Main()
<-exitChan
nsqadmin.Exit()
}
func testNsqLookupNsqdNodesChange(t *testing.T, useFakeLeadership bool) {
if testing.Verbose() {
SetCoordLogger(&levellogger.GLogger{}, levellogger.LOG_INFO)
glog.SetFlags(0, "", "", true, true, 1)
glog.StartWorker(time.Second)
} else {
SetCoordLogger(newTestLogger(t), levellogger.LOG_DEBUG)
}
idList := []string{"id1", "id2", "id3", "id4", "id5"}
lookupCoord1, nodeInfoList := prepareCluster(t, idList, useFakeLeadership)
for _, n := range nodeInfoList {
defer os.RemoveAll(n.dataPath)
defer n.localNsqd.Exit()
defer n.nsqdCoord.Stop()
}
topic := "test-nsqlookup-topic-unit-test"
lookupLeadership := lookupCoord1.leadership
lookupCoord1.DeleteTopic(topic, "**")
topic3 := topic + topic
lookupCoord1.DeleteTopic(topic3, "**")
time.Sleep(time.Second)
defer func() {
lookupCoord1.DeleteTopic(topic, "**")
lookupCoord1.DeleteTopic(topic3, "**")
time.Sleep(time.Second * 3)
lookupCoord1.Stop()
}()
// test new topic create
err := lookupCoord1.CreateTopic(topic, TopicMetaInfo{2, 2, 0, 0, 0, 0})
test.Nil(t, err)
waitClusterStable(lookupCoord1, time.Second*3)
pmeta, _, err := lookupLeadership.GetTopicMetaInfo(topic)
pn := pmeta.PartitionNum
test.Nil(t, err)
test.Equal(t, pn, 2)
t0, err := lookupLeadership.GetTopicInfo(topic, 0)
test.Nil(t, err)
t1, err := lookupLeadership.GetTopicInfo(topic, 1)
test.Nil(t, err)
test.Equal(t, len(t0.ISR), 2)
test.Equal(t, len(t1.ISR), 2)
t.Log(t0)
t.Log(t1)
test.NotEqual(t, t0.Leader, t1.Leader)
t0LeaderCoord := nodeInfoList[t0.Leader].nsqdCoord
test.NotNil(t, t0LeaderCoord)
tc0, coordErr := t0LeaderCoord.getTopicCoord(topic, 0)
test.Nil(t, coordErr)
test.Equal(t, tc0.topicInfo.Leader, t0.Leader)
test.Equal(t, len(tc0.topicInfo.ISR), 2)
t1LeaderCoord := nodeInfoList[t1.Leader].nsqdCoord
test.NotNil(t, t1LeaderCoord)
tc1, coordErr := t1LeaderCoord.getTopicCoord(topic, 1)
test.Nil(t, coordErr)
test.Equal(t, tc1.topicInfo.Leader, t1.Leader)
test.Equal(t, len(tc1.topicInfo.ISR), 2)
coordLog.Warningf("============= begin test isr node failed ====")
// test isr node lost
lostNodeID := t0.ISR[1]
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 1)
nodeInfoList[lostNodeID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
t0, err = lookupLeadership.GetTopicInfo(topic, 0)
if len(t0.ISR) < t0.Replica {
waitClusterStable(lookupCoord1, time.Second*3)
}
t0, err = lookupLeadership.GetTopicInfo(topic, 0)
test.Nil(t, err)
test.Equal(t, FindSlice(t0.ISR, lostNodeID) == -1, true)
test.Equal(t, len(t0.ISR), t0.Replica)
test.Equal(t, t0.Leader, t0.ISR[0])
// clear topic info on failed node, test the reload for failed node
nodeInfoList[lostNodeID].nsqdCoord.topicCoords = make(map[string]map[int]*TopicCoordinator)
// test new catchup and new isr
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 0)
nodeInfoList[lostNodeID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
if len(t0.ISR) < t0.Replica {
waitClusterStable(lookupCoord1, time.Second*3)
}
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
test.Equal(t, len(t0.CatchupList), 0)
test.Equal(t, len(t0.ISR) >= t0.Replica, true)
test.Equal(t, len(tc0.topicInfo.ISR), len(t0.ISR))
test.Equal(t, t0.Leader, t0.ISR[0])
lookupCoord1.triggerCheckTopics("", 0, time.Second)
time.Sleep(time.Second)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
// should remove the unnecessary node
test.Equal(t, len(t0.ISR), t0.Replica)
coordLog.Warningf("============= begin test leader failed ====")
// test leader node lost
lostNodeID = t0.Leader
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 1)
nodeInfoList[lostNodeID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
if len(t0.ISR) < t0.Replica {
waitClusterStable(lookupCoord1, time.Second*3)
}
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
t.Log(t0)
test.Equal(t, t0.Replica, len(t0.ISR))
test.Equal(t, t0.Leader, t0.ISR[0])
test.NotEqual(t, t0.Leader, lostNodeID)
//test.Equal(t, len(t0.CatchupList), 1)
test.Equal(t, FindSlice(t0.ISR, lostNodeID) == -1, true)
t0LeaderCoord = nodeInfoList[t0.Leader].nsqdCoord
test.NotNil(t, t0LeaderCoord)
tc0, coordErr = t0LeaderCoord.getTopicCoord(topic, 0)
test.Nil(t, coordErr)
test.Equal(t, len(tc0.topicInfo.ISR), len(t0.ISR))
test.Equal(t, tc0.topicInfo.Leader, t0.Leader)
// test lost leader node rejoin
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 0)
nodeInfoList[lostNodeID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
if len(t0.ISR) < t0.Replica {
waitClusterStable(lookupCoord1, time.Second*3)
}
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
t.Log(t0)
test.Equal(t, len(t0.CatchupList), 0)
test.Equal(t, len(t0.ISR) >= t0.Replica, true)
t0LeaderCoord = nodeInfoList[t0.Leader].nsqdCoord
test.NotNil(t, t0LeaderCoord)
tc0, coordErr = t0LeaderCoord.getTopicCoord(topic, 0)
test.Nil(t, coordErr)
test.Equal(t, len(tc0.topicInfo.ISR), len(t0.ISR))
test.Equal(t, tc0.topicInfo.Leader, t0.Leader)
waitClusterStable(lookupCoord1, time.Second*3)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
// should remove the unnecessary node
test.Equal(t, len(t0.ISR), t0.Replica)
// test old leader failed and begin elect new and then new leader failed
coordLog.Warningf("============= begin test old leader failed and then new leader failed ====")
lostNodeID = t0.Leader
lostISRID := t0.ISR[1]
if lostISRID == lostNodeID {
lostISRID = t0.ISR[0]
}
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 1)
nodeInfoList[lostNodeID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
time.Sleep(time.Millisecond)
atomic.StoreInt32(&nodeInfoList[lostISRID].nsqdCoord.stopping, 1)
nodeInfoList[lostISRID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostISRID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 0)
atomic.StoreInt32(&nodeInfoList[lostISRID].nsqdCoord.stopping, 0)
nodeInfoList[lostNodeID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
nodeInfoList[lostISRID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostISRID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*3)
waitClusterStable(lookupCoord1, time.Second*5)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
test.Equal(t, true, len(t0.ISR) >= t0.Replica)
test.Equal(t, t0.Leader == t0.ISR[0] || t0.Leader == t0.ISR[1], true)
t0LeaderCoord = nodeInfoList[t0.Leader].nsqdCoord
test.NotNil(t, t0LeaderCoord)
tc0, coordErr = t0LeaderCoord.getTopicCoord(topic, 0)
test.Nil(t, coordErr)
test.Equal(t, len(tc0.topicInfo.ISR), len(t0.ISR))
test.Equal(t, tc0.topicInfo.Leader, t0.Leader)
waitClusterStable(lookupCoord1, time.Second*5)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
// should remove the unnecessary node
test.Equal(t, t0.Replica, len(t0.ISR))
// test join isr timeout
lostNodeID = t1.ISR[1]
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 1)
nodeInfoList[lostNodeID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*5)
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 0)
nodeInfoList[lostNodeID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*5)
// with only 2 replica, the isr join fail should not change the isr list
nodeInfoList[lostNodeID].nsqdCoord.rpcServer.toggleDisableRpcTest(true)
waitClusterStable(lookupCoord1, time.Second*10)
t1, _ = lookupLeadership.GetTopicInfo(topic, 1)
test.Equal(t, true, len(t1.ISR)+len(t1.CatchupList) >= t1.Replica)
test.Equal(t, t1.Leader == t1.ISR[0] || t1.Leader == t1.ISR[1], true)
nodeInfoList[lostNodeID].nsqdCoord.rpcServer.toggleDisableRpcTest(false)
waitClusterStable(lookupCoord1, time.Second*5)
// test new topic create
coordLog.Warningf("============= begin test 3 replicas ====")
err = lookupCoord1.CreateTopic(topic3, TopicMetaInfo{1, 3, 0, 0, 0, 0})
test.Nil(t, err)
waitClusterStable(lookupCoord1, time.Second*5)
// with 3 replica, the isr join timeout will change the isr list if the isr has the quorum nodes
t3, err := lookupLeadership.GetTopicInfo(topic3, 0)
test.Nil(t, err)
test.Equal(t, len(t3.ISR), t3.Replica)
lostNodeID = t3.ISR[1]
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 1)
nodeInfoList[lostNodeID].nsqdCoord.leadership.UnregisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*5)
atomic.StoreInt32(&nodeInfoList[lostNodeID].nsqdCoord.stopping, 0)
nodeInfoList[lostNodeID].nsqdCoord.leadership.RegisterNsqd(nodeInfoList[lostNodeID].nodeInfo)
waitClusterStable(lookupCoord1, time.Second*5)
nodeInfoList[lostNodeID].nsqdCoord.rpcServer.toggleDisableRpcTest(true)
waitClusterStable(lookupCoord1, time.Second*5)
t3, _ = lookupLeadership.GetTopicInfo(topic3, 0)
test.Equal(t, true, len(t3.ISR) >= t3.Replica-1)
test.Equal(t, true, len(t3.ISR) <= t3.Replica)
test.Equal(t, t3.Leader == t3.ISR[0] || t3.Leader == t3.ISR[1], true)
nodeInfoList[lostNodeID].nsqdCoord.rpcServer.toggleDisableRpcTest(false)
waitClusterStable(lookupCoord1, time.Second*5)
glog.Flush()
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
test.Equal(t, true, len(t0.ISR) >= t0.Replica)
t1, _ = lookupLeadership.GetTopicInfo(topic, 1)
test.Equal(t, true, len(t1.ISR) >= t0.Replica)
// before migrate really start, the isr should not reach the replica factor
// however, catch up may start early while check leadership or enable topic write
t3, _ = lookupLeadership.GetTopicInfo(topic3, 0)
test.Equal(t, true, len(t3.ISR)+len(t3.CatchupList) >= t3.Replica)
t0IsrNum := 2
t1IsrNum := 2
coordLog.Warningf("========== begin test quit ====")
quitList := make([]*NsqdCoordinator, 0)
quitList = append(quitList, nodeInfoList[t0.Leader].nsqdCoord)
if t1.Leader != t0.Leader {
quitList = append(quitList, nodeInfoList[t1.Leader].nsqdCoord)
}
if t3.Leader != t0.Leader && t3.Leader != t1.Leader {
quitList = append(quitList, nodeInfoList[t3.Leader].nsqdCoord)
}
for id, n := range nodeInfoList {
if id == t0.Leader || id == t1.Leader || id == t3.Leader {
continue
}
quitList = append(quitList, n.nsqdCoord)
}
test.Equal(t, len(nodeInfoList), len(quitList))
for _, nsqdCoord := range quitList {
failedID := nsqdCoord.myNode.GetID()
delete(nodeInfoList, failedID)
nsqdCoord.Stop()
if t0IsrNum > 1 {
if FindSlice(t0.ISR, failedID) != -1 {
t0IsrNum--
}
}
if t1IsrNum > 1 {
if FindSlice(t1.ISR, failedID) != -1 {
t1IsrNum--
}
}
waitClusterStable(lookupCoord1, time.Second*5)
t0, _ = lookupLeadership.GetTopicInfo(topic, 0)
// we have no failed node in isr or we got the last failed node leaving in isr.
t.Log(t0)
test.Equal(t, FindSlice(t0.ISR, failedID) == -1 || (len(t0.ISR) == 1 && t0.ISR[0] == failedID), true)
test.Equal(t, true, len(t0.ISR) >= t0IsrNum)
t1, _ = lookupLeadership.GetTopicInfo(topic, 1)
t.Log(t1)
test.Equal(t, FindSlice(t1.ISR, failedID) == -1 || (len(t1.ISR) == 1 && t1.ISR[0] == failedID), true)
test.Equal(t, true, len(t1.ISR) >= t1IsrNum)
t3, _ = lookupLeadership.GetTopicInfo(topic3, 0)
t.Log(t3)
test.Equal(t, FindSlice(t3.ISR, failedID) == -1 || (len(t3.ISR) == 1 && t3.ISR[0] == failedID), true)
}
}