func TestNsqdRPCClient(t *testing.T) {
SetCoordLogger(newTestLogger(t), 2)
tmpDir, err := ioutil.TempDir("", fmt.Sprintf("nsq-test-%d", time.Now().UnixNano()))
if err != nil {
panic(err)
}
defer os.RemoveAll(tmpDir)
nsqdCoord := startNsqdCoord(t, "0", tmpDir, "", nil, true)
nsqdCoord.Start()
defer nsqdCoord.Stop()
time.Sleep(time.Second * 2)
client, err := NewNsqdRpcClient(nsqdCoord.rpcServer.rpcServer.Listener.ListenAddr().String(), time.Second)
test.Nil(t, err)
_, err = client.CallWithRetry("TestRpcCallNotExist", "req")
test.NotNil(t, err)
coordErr := client.CallRpcTestCoordErr("coorderr")
test.NotNil(t, coordErr)
test.NotEqual(t, coordErr.ErrType, CoordNetErr)
test.Equal(t, coordErr.ErrMsg, "coorderr")
test.Equal(t, coordErr.ErrCode, RpcCommonErr)
test.Equal(t, coordErr.ErrType, CoordCommonErr)
rsp, rpcErr := client.CallRpcTest("reqdata")
test.NotNil(t, rpcErr)
test.Equal(t, rsp, "reqdata")
test.Equal(t, rpcErr.ErrCode, RpcNoErr)
test.Equal(t, rpcErr.ErrMsg, "reqdata")
test.Equal(t, rpcErr.ErrType, CoordCommonErr)
timeoutErr := client.CallRpcTesttimeout("reqdata")
test.NotNil(t, timeoutErr)
test.Equal(t, timeoutErr.(*gorpc.ClientError).Timeout, true)
time.Sleep(time.Second * 3)
client.Close()
}
func TestGetTopic(t *testing.T) {
opts := NewOptions()
opts.Logger = newTestLogger(t)
_, _, nsqd := mustStartNSQD(opts)
defer os.RemoveAll(opts.DataPath)
defer nsqd.Exit()
topic1 := nsqd.GetTopic("test", 0)
test.NotNil(t, topic1)
test.Equal(t, "test", topic1.GetTopicName())
topic2 := nsqd.GetTopic("test", 0)
test.Equal(t, topic1, topic2)
topic3 := nsqd.GetTopic("test2", 1)
test.Equal(t, "test2", topic3.GetTopicName())
test.NotEqual(t, topic2, topic3)
topic1_1 := nsqd.GetTopicIgnPart("test")
test.Equal(t, "test", topic1_1.GetTopicName())
test.Equal(t, 0, topic1_1.GetTopicPart())
topic3_1 := nsqd.GetTopicIgnPart("test2")
test.Equal(t, "test2", topic3_1.GetTopicName())
test.Equal(t, 1, topic3_1.GetTopicPart())
}
func TestTopicPutChannelWait(t *testing.T) {
opts := NewOptions()
opts.Logger = newTestLogger(t)
_, _, nsqd := mustStartNSQD(opts)
defer os.RemoveAll(opts.DataPath)
defer nsqd.Exit()
topic := nsqd.GetTopic("test", 0)
topic.dynamicConf.AutoCommit = 1
topic.dynamicConf.SyncEvery = 10
channel := topic.GetChannel("ch")
test.NotNil(t, channel)
msg := NewMessage(0, []byte("aaaaaaaaaaaaaaaaaaaaaaaaaaa"))
for i := 0; i <= 10; i++ {
msg.ID = 0
topic.PutMessage(msg)
}
topic.ForceFlush()
test.Equal(t, topic.backend.GetQueueReadEnd(), topic.backend.GetQueueWriteEnd())
test.Equal(t, topic.backend.GetQueueReadEnd(), channel.GetChannelEnd())
for i := 0; i <= 10; i++ {
select {
case outMsg := <-channel.clientMsgChan:
test.Equal(t, msg.Body, outMsg.Body)
channel.ConfirmBackendQueue(outMsg)
case <-time.After(time.Second):
t.Fatalf("should read message in channel")
}
}
test.Equal(t, true, channel.IsWaitingMoreData())
test.Equal(t, topic.backend.GetQueueReadEnd(), channel.GetChannelEnd())
msg.ID = 0
topic.PutMessage(msg)
test.Equal(t, false, channel.IsWaitingMoreData())
test.Equal(t, topic.backend.GetQueueReadEnd(), topic.backend.GetQueueWriteEnd())
test.Equal(t, topic.backend.GetQueueReadEnd(), channel.GetChannelEnd())
select {
case outMsg := <-channel.clientMsgChan:
test.Equal(t, msg.Body, outMsg.Body)
channel.ConfirmBackendQueue(outMsg)
case <-time.After(time.Second):
t.Fatalf("should read the message in channel")
}
test.Equal(t, true, channel.IsWaitingMoreData())
msg.ID = 0
topic.PutMessage(msg)
test.Equal(t, false, channel.IsWaitingMoreData())
test.Equal(t, topic.backend.GetQueueReadEnd(), topic.backend.GetQueueWriteEnd())
test.Equal(t, topic.backend.GetQueueReadEnd(), channel.GetChannelEnd())
msg.ID = 0
topic.PutMessage(msg)
test.NotEqual(t, topic.backend.GetQueueReadEnd(), topic.backend.GetQueueWriteEnd())
test.Equal(t, topic.backend.GetQueueReadEnd(), channel.GetChannelEnd())
}
func TestHTTPpub(t *testing.T) {
opts := nsqd.NewOptions()
opts.LogLevel = 2
opts.Logger = newTestLogger(t)
//opts.Logger = &levellogger.GLogger{}
tcpAddr, httpAddr, nsqd, nsqdServer := mustStartNSQD(opts)
defer os.RemoveAll(opts.DataPath)
defer nsqdServer.Exit()
topicName := "test_http_pub" + strconv.Itoa(int(time.Now().Unix()))
_ = nsqd.GetTopicIgnPart(topicName)
conn, err := mustConnectNSQD(tcpAddr)
test.Equal(t, err, nil)
identify(t, conn, nil, frameTypeResponse)
sub(t, conn, topicName, "ch")
buf := bytes.NewBuffer([]byte("test message"))
url := fmt.Sprintf("http://%s/pub?topic=%s", httpAddr, topicName)
resp, err := http.Post(url, "application/octet-stream", buf)
test.Equal(t, err, nil)
defer resp.Body.Close()
body, _ := ioutil.ReadAll(resp.Body)
test.Equal(t, string(body), "OK")
time.Sleep(5 * time.Millisecond)
_, err = nsq.Ready(1).WriteTo(conn)
test.Equal(t, err, nil)
// sleep to allow the RDY state to take effect
time.Sleep(50 * time.Millisecond)
for {
resp, _ := nsq.ReadResponse(conn)
frameType, data, err := nsq.UnpackResponse(resp)
test.Nil(t, err)
test.NotEqual(t, frameTypeError, frameType)
if frameType == frameTypeResponse {
t.Logf("got response data: %v", string(data))
continue
}
msgOut, err := nsq.DecodeMessage(data)
test.Equal(t, []byte("test message"), msgOut.Body)
_, err = nsq.Finish(msgOut.ID).WriteTo(conn)
test.Nil(t, err)
break
}
conn.Close()
}
func TestHTTPpubpartition(t *testing.T) {
opts := nsqd.NewOptions()
opts.LogLevel = 2
opts.Logger = newTestLogger(t)
//opts.Logger = &levellogger.GLogger{}
_, httpAddr, nsqd, nsqdServer := mustStartNSQD(opts)
defer os.RemoveAll(opts.DataPath)
defer nsqdServer.Exit()
topicName := "test_http_pub_partition" + strconv.Itoa(int(time.Now().Unix()))
_ = nsqd.GetTopicIgnPart(topicName)
buf := bytes.NewBuffer([]byte("test message"))
// should failed pub to not exist partition
url := fmt.Sprintf("http://%s/pub?topic=%s&partition=2", httpAddr, topicName)
resp, err := http.Post(url, "application/octet-stream", buf)
test.Equal(t, err, nil)
defer resp.Body.Close()
body, _ := ioutil.ReadAll(resp.Body)
test.NotEqual(t, string(body), "OK")
time.Sleep(5 * time.Millisecond)
}
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)
}
}
func TestNsqLookupMovePartition(t *testing.T) {
if testing.Verbose() {
SetCoordLogger(&levellogger.GLogger{}, levellogger.LOG_WARN)
glog.SetFlags(0, "", "", true, true, 1)
glog.StartWorker(time.Second)
} else {
SetCoordLogger(newTestLogger(t), levellogger.LOG_DEBUG)
}
idList := []string{"id1", "id2", "id3", "id4", "id5"}
lookupCoord, nodeInfoList := prepareCluster(t, idList, false)
for _, n := range nodeInfoList {
defer os.RemoveAll(n.dataPath)
defer n.localNsqd.Exit()
defer n.nsqdCoord.Stop()
}
topic_p1_r1 := "test-nsqlookup-topic-unit-test-move-p1-r1"
topic_p2_r2 := "test-nsqlookup-topic-unit-test-move-p2-r2"
lookupLeadership := lookupCoord.leadership
checkDeleteErr(t, lookupCoord.DeleteTopic(topic_p1_r1, "**"))
checkDeleteErr(t, lookupCoord.DeleteTopic(topic_p2_r2, "**"))
time.Sleep(time.Second * 3)
defer func() {
checkDeleteErr(t, lookupCoord.DeleteTopic(topic_p1_r1, "**"))
checkDeleteErr(t, lookupCoord.DeleteTopic(topic_p2_r2, "**"))
time.Sleep(time.Second * 3)
lookupCoord.Stop()
}()
// test new topic create
err := lookupCoord.CreateTopic(topic_p1_r1, TopicMetaInfo{1, 1, 0, 0, 0, 0})
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
err = lookupCoord.CreateTopic(topic_p2_r2, TopicMetaInfo{2, 2, 0, 0, 0, 0})
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
lookupCoord.triggerCheckTopics("", 0, 0)
waitClusterStable(lookupCoord, time.Second*3)
// test move leader to other isr;
// test move leader to other catchup;
// test move non-leader to other node;
t0, err := lookupLeadership.GetTopicInfo(topic_p1_r1, 0)
test.Nil(t, err)
test.Equal(t, len(t0.ISR), 1)
// move p1_r1 leader to other node
toNode := ""
for _, node := range nodeInfoList {
if node.nodeInfo.GetID() == t0.Leader {
continue
}
toNode = node.nodeInfo.GetID()
break
}
lookupCoord.triggerCheckTopics("", 0, 0)
time.Sleep(time.Second)
err = lookupCoord.MoveTopicPartitionDataByManual(topic_p1_r1, 0, true, t0.Leader, toNode)
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
t0, err = lookupLeadership.GetTopicInfo(topic_p1_r1, 0)
test.Nil(t, err)
// it may be two nodes in isr if the moved leader rejoin as isr
test.Equal(t, len(t0.ISR) >= 1, true)
test.Equal(t, t0.Leader, toNode)
t0, err = lookupLeadership.GetTopicInfo(topic_p2_r2, 0)
test.Nil(t, err)
test.Equal(t, len(t0.ISR), 2)
toNode = ""
for _, nid := range t0.ISR {
if nid == t0.Leader {
continue
}
toNode = nid
break
}
waitClusterStable(lookupCoord, time.Second*3)
// move leader to other isr node
oldLeader := t0.Leader
err = lookupCoord.MoveTopicPartitionDataByManual(topic_p2_r2, 0, true, t0.Leader, toNode)
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
t0, err = lookupLeadership.GetTopicInfo(topic_p2_r2, 0)
test.Nil(t, err)
test.Equal(t, len(t0.ISR) >= 2, true)
test.NotEqual(t, t0.Leader, oldLeader)
test.Equal(t, t0.Leader, toNode)
// move leader to other non-isr node
toNode = ""
for _, node := range nodeInfoList {
if FindSlice(t0.ISR, node.nodeInfo.GetID()) != -1 {
continue
}
// check other partition
t1, err := lookupLeadership.GetTopicInfo(topic_p2_r2, 1)
if err == nil {
if FindSlice(t1.ISR, node.nodeInfo.GetID()) != -1 {
continue
}
}
toNode = node.nodeInfo.GetID()
break
}
lookupCoord.triggerCheckTopics("", 0, 0)
time.Sleep(time.Second)
err = lookupCoord.MoveTopicPartitionDataByManual(topic_p2_r2, 0, true, t0.Leader, toNode)
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
t0, err = lookupLeadership.GetTopicInfo(topic_p2_r2, 0)
test.Nil(t, err)
test.Equal(t, t0.Leader, toNode)
// move non-leader to other non-isr node
toNode = ""
toNodeInvalid := ""
fromNode := ""
for _, nid := range t0.ISR {
if nid != t0.Leader {
fromNode = nid
}
}
for _, node := range nodeInfoList {
if FindSlice(t0.ISR, node.nodeInfo.GetID()) != -1 {
continue
}
// check other partition
t1, err := lookupLeadership.GetTopicInfo(topic_p2_r2, 1)
if err == nil {
toNodeInvalid = t1.Leader
if FindSlice(t1.ISR, node.nodeInfo.GetID()) != -1 {
continue
}
}
toNode = node.nodeInfo.GetID()
break
}
lookupCoord.triggerCheckTopics("", 0, 0)
time.Sleep(time.Second)
err = lookupCoord.MoveTopicPartitionDataByManual(topic_p2_r2, 0, false, fromNode, toNodeInvalid)
test.NotNil(t, err)
test.Equal(t, ErrNodeIsExcludedForTopicData, err)
lookupCoord.triggerCheckTopics("", 0, 0)
time.Sleep(time.Second)
err = lookupCoord.MoveTopicPartitionDataByManual(topic_p2_r2, 0, false, fromNode, toNode)
test.Nil(t, err)
waitClusterStable(lookupCoord, time.Second*3)
t0, err = lookupLeadership.GetTopicInfo(topic_p2_r2, 0)
test.Nil(t, err)
test.Equal(t, FindSlice(t0.ISR, toNode) != -1, true)
test.Equal(t, -1, FindSlice(t0.ISR, fromNode))
}