func (p *protocolV2) FIN(client *nsqd.ClientV2, params [][]byte) ([]byte, error) {
state := atomic.LoadInt32(&client.State)
if state != stateSubscribed && state != stateClosing {
nsqd.NsqLogger().LogWarningf("[%s] command in wrong state: %v", client, state)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "cannot FIN in current state")
}
if len(params) < 2 {
nsqd.NsqLogger().LogDebugf("FIN error params: %v", params)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "FIN insufficient number of params")
}
id, err := getFullMessageID(params[1])
if err != nil {
nsqd.NsqLogger().LogDebugf("FIN error: %v, %v", params[1], err)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, err.Error())
}
msgID := nsqd.GetMessageIDFromFullMsgID(*id)
if int64(msgID) <= 0 {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "Invalid Message ID")
}
if client.Channel == nil {
nsqd.NsqLogger().LogDebugf("FIN error no channel: %v", msgID)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "No channel")
}
if !p.ctx.checkForMasterWrite(client.Channel.GetTopicName(), client.Channel.GetTopicPart()) {
nsqd.NsqLogger().Logf("topic %v fin message failed for not leader", client.Channel.GetTopicName())
return nil, protocol.NewFatalClientErr(nil, FailedOnNotLeader, "")
}
err = p.ctx.FinishMessage(client.Channel, client.ID, client.String(), msgID)
if err != nil {
client.IncrSubError(int64(1))
nsqd.NsqLogger().LogDebugf("FIN error : %v, err: %v, channel: %v, topic: %v", msgID,
err, client.Channel.GetName(), client.Channel.GetTopicName())
if clusterErr, ok := err.(*consistence.CommonCoordErr); ok {
if !clusterErr.IsLocalErr() {
return nil, protocol.NewFatalClientErr(err, FailedOnNotWritable, "")
}
}
return nil, protocol.NewClientErr(err, "E_FIN_FAILED",
fmt.Sprintf("FIN %v failed %s", *id, err.Error()))
}
client.FinishedMessage()
return nil, nil
}
func (p *protocolV2) REQ(client *nsqd.ClientV2, params [][]byte) ([]byte, error) {
state := atomic.LoadInt32(&client.State)
if state != stateSubscribed && state != stateClosing {
nsqd.NsqLogger().LogWarningf("[%s] command in wrong state: %v", client, state)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "cannot REQ in current state")
}
if len(params) < 3 {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "REQ insufficient number of params")
}
id, err := getFullMessageID(params[1])
if err != nil {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, err.Error())
}
timeoutMs, err := protocol.ByteToBase10(params[2])
if err != nil {
return nil, protocol.NewFatalClientErr(err, E_INVALID,
fmt.Sprintf("REQ could not parse timeout %s", params[2]))
}
timeoutDuration := time.Duration(timeoutMs) * time.Millisecond
if timeoutDuration < 0 || timeoutDuration > p.ctx.getOpts().MaxReqTimeout {
return nil, protocol.NewFatalClientErr(nil, E_INVALID,
fmt.Sprintf("REQ timeout %v out of range 0-%v", timeoutDuration, p.ctx.getOpts().MaxReqTimeout))
}
if client.Channel == nil {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "No channel")
}
err = client.Channel.RequeueMessage(client.ID, client.String(), nsqd.GetMessageIDFromFullMsgID(*id), timeoutDuration, true)
if err != nil {
client.IncrSubError(int64(1))
return nil, protocol.NewClientErr(err, "E_REQ_FAILED",
fmt.Sprintf("REQ %v failed %s", *id, err.Error()))
}
client.RequeuedMessage(timeoutDuration > 0)
return nil, nil
}
func (p *protocolV2) internalSUB(client *nsqd.ClientV2, params [][]byte, enableTrace bool,
ordered bool, startFrom *ConsumeOffset) ([]byte, error) {
state := atomic.LoadInt32(&client.State)
if state != stateInit {
nsqd.NsqLogger().LogWarningf("[%s] command in wrong state: %v", client, state)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "cannot SUB in current state")
}
if client.HeartbeatInterval <= 0 {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "cannot SUB with heartbeats disabled")
}
if len(params) < 3 {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, "SUB insufficient number of parameters")
}
topicName := string(params[1])
if !protocol.IsValidTopicName(topicName) {
return nil, protocol.NewFatalClientErr(nil, "E_BAD_TOPIC",
fmt.Sprintf("SUB topic name %q is not valid", topicName))
}
partition := -1
channelName := ""
var err error
channelName = string(params[2])
if !protocol.IsValidChannelName(channelName) {
return nil, protocol.NewFatalClientErr(nil, "E_BAD_CHANNEL",
fmt.Sprintf("SUB channel name %q is not valid", channelName))
}
if len(params) == 4 {
partition, err = strconv.Atoi(string(params[3]))
if err != nil {
return nil, protocol.NewFatalClientErr(nil, "E_BAD_PARTITION",
fmt.Sprintf("topic partition is not valid: %v", err))
}
}
if err = p.CheckAuth(client, "SUB", topicName, channelName); err != nil {
return nil, err
}
if partition == -1 {
partition = p.ctx.getDefaultPartition(topicName)
}
topic, err := p.ctx.getExistingTopic(topicName, partition)
if err != nil {
nsqd.NsqLogger().Logf("sub to not existing topic: %v, err:%v", topicName, err.Error())
return nil, protocol.NewFatalClientErr(nil, E_TOPIC_NOT_EXIST, "")
}
if !p.ctx.checkForMasterWrite(topicName, partition) {
nsqd.NsqLogger().Logf("sub failed on not leader: %v-%v, remote is : %v", topicName, partition, client.RemoteAddr())
// we need disable topic here to trigger a notify, maybe we failed to notify lookup last time.
topic.DisableForSlave()
return nil, protocol.NewFatalClientErr(nil, FailedOnNotLeader, "")
}
channel := topic.GetChannel(channelName)
err = channel.AddClient(client.ID, client)
if err != nil {
nsqd.NsqLogger().Logf("sub failed to add client: %v, %v", client, err)
return nil, protocol.NewFatalClientErr(nil, FailedOnNotWritable, "")
}
atomic.StoreInt32(&client.State, stateSubscribed)
client.Channel = channel
if enableTrace {
nsqd.NsqLogger().Logf("sub channel %v with trace enabled, remote is : %v", channelName, client.RemoteAddr())
}
if ordered {
if atomic.LoadInt32(&client.SampleRate) != 0 {
nsqd.NsqLogger().Errorf("%v", ErrOrderChannelOnSampleRate)
return nil, protocol.NewFatalClientErr(nil, E_INVALID, ErrOrderChannelOnSampleRate.Error())
}
channel.SetOrdered(true)
}
if startFrom != nil {
cnt := channel.GetClientsCount()
if cnt > 1 {
nsqd.NsqLogger().LogDebugf("the consume offset: %v can only be set by the first client: %v", startFrom, cnt)
} else {
queueOffset, cnt, err := p.ctx.SetChannelOffset(channel, startFrom, false)
if err != nil {
return nil, protocol.NewFatalClientErr(nil, E_INVALID, err.Error())
}
nsqd.NsqLogger().Logf("set the channel offset: %v (actual set : %v:%v), by client:%v, %v",
startFrom, queueOffset, cnt, client.String(), client.UserAgent)
}
}
client.EnableTrace = enableTrace
// update message pump
client.SubEventChan <- channel
return okBytes, nil
}
func (p *protocolV2) messagePump(client *nsqd.ClientV2, startedChan chan bool,
stoppedChan chan bool) {
var err error
var buf bytes.Buffer
var clientMsgChan chan *nsqd.Message
var subChannel *nsqd.Channel
// NOTE: `flusherChan` is used to bound message latency for
// the pathological case of a channel on a low volume topic
// with >1 clients having >1 RDY counts
var flusherChan <-chan time.Time
var sampleRate int32
subEventChan := client.SubEventChan
identifyEventChan := client.IdentifyEventChan
outputBufferTicker := time.NewTicker(client.OutputBufferTimeout)
heartbeatTicker := time.NewTicker(client.HeartbeatInterval)
heartbeatChan := heartbeatTicker.C
heartbeatFailedCnt := 0
msgTimeout := client.MsgTimeout
// v2 opportunistically buffers data to clients to reduce write system calls
// we force flush in two cases:
// 1. when the client is not ready to receive messages
// 2. we're buffered and the channel has nothing left to send us
// (ie. we would block in this loop anyway)
//
flushed := true
// signal to the goroutine that started the messagePump
// that we've started up
close(startedChan)
for {
if subChannel == nil || !client.IsReadyForMessages() {
// the client is not ready to receive messages...
clientMsgChan = nil
flusherChan = nil
// force flush
client.LockWrite()
err = client.Flush()
client.UnlockWrite()
if err != nil {
goto exit
}
flushed = true
} else if flushed {
// last iteration we flushed...
// do not select on the flusher ticker channel
clientMsgChan = subChannel.GetClientMsgChan()
flusherChan = nil
} else {
// we're buffered (if there isn't any more data we should flush)...
// select on the flusher ticker channel, too
clientMsgChan = subChannel.GetClientMsgChan()
flusherChan = outputBufferTicker.C
}
select {
case <-client.ExitChan:
goto exit
case <-flusherChan:
// if this case wins, we're either starved
// or we won the race between other channels...
// in either case, force flush
client.LockWrite()
err = client.Flush()
client.UnlockWrite()
if err != nil {
goto exit
}
flushed = true
case <-client.ReadyStateChan:
case subChannel = <-subEventChan:
// you can't SUB anymore
nsqd.NsqLogger().Logf("client %v sub to channel: %v", client.ID,
subChannel.GetName())
subEventChan = nil
case identifyData := <-identifyEventChan:
// you can't IDENTIFY anymore
identifyEventChan = nil
outputBufferTicker.Stop()
if identifyData.OutputBufferTimeout > 0 {
outputBufferTicker = time.NewTicker(identifyData.OutputBufferTimeout)
}
heartbeatTicker.Stop()
heartbeatChan = nil
if identifyData.HeartbeatInterval > 0 {
heartbeatTicker = time.NewTicker(identifyData.HeartbeatInterval)
heartbeatChan = heartbeatTicker.C
}
if identifyData.SampleRate > 0 {
sampleRate = identifyData.SampleRate
}
msgTimeout = identifyData.MsgTimeout
case <-heartbeatChan:
if subChannel != nil && client.IsReadyForMessages() {
// try wake up the channel
subChannel.TryWakeupRead()
}
err = Send(client, frameTypeResponse, heartbeatBytes)
nsqd.NsqLogger().LogDebugf("PROTOCOL(V2): [%s] send heartbeat", client)
if err != nil {
heartbeatFailedCnt++
nsqd.NsqLogger().LogWarningf("PROTOCOL(V2): [%s] send heartbeat failed %v times, %v", client, heartbeatFailedCnt, err)
if heartbeatFailedCnt > 2 {
goto exit
}
} else {
heartbeatFailedCnt = 0
}
case msg, ok := <-clientMsgChan:
if !ok {
goto exit
}
if sampleRate > 0 && rand.Int31n(100) > sampleRate {
// FIN automatically, all message will not wait to confirm if not sending,
// and the reader keep moving forward.
offset, _, _, _ := subChannel.ConfirmBackendQueue(msg)
// TODO: sync to replica nodes.
_ = offset
continue
}
// avoid re-send some confirmed message,
// this may happen while the channel reader is reset to old position
// due to some retry or leader change.
if subChannel.IsConfirmed(msg) {
continue
}
subChannel.StartInFlightTimeout(msg, client.ID, client.String(), msgTimeout)
client.SendingMessage()
err = SendMessage(client, msg, &buf, subChannel.IsOrdered())
if err != nil {
goto exit
}
flushed = false
}
}
exit:
nsqd.NsqLogger().LogDebugf("PROTOCOL(V2): [%s] exiting messagePump", client)
heartbeatTicker.Stop()
outputBufferTicker.Stop()
if err != nil {
nsqd.NsqLogger().Logf("PROTOCOL(V2): [%s] messagePump error - %s", client, err)
}
close(stoppedChan)
}
