Github user tillrohrmann commented on a diff in the pull request:
https://github.com/apache/flink/pull/328#discussion_r23523589
--- Diff:
flink-runtime/src/main/scala/org/apache/flink/runtime/taskmanager/TaskManager.scala
---
@@ -175,62 +175,79 @@ import scala.collection.JavaConverters._
}
private def tryJobManagerRegistration(): Unit = {
- registrationAttempts = 0
- import context.dispatcher
- registrationScheduler = Some(context.system.scheduler.schedule(
- TaskManager.REGISTRATION_DELAY, TaskManager.REGISTRATION_INTERVAL,
- self, RegisterAtJobManager))
+ registrationDuration = 0 seconds
+
+ registered = false
+
+ context.system.scheduler.scheduleOnce(registrationDelay, self,
RegisterAtJobManager)
}
override def receiveWithLogMessages: Receive = {
case RegisterAtJobManager => {
- registrationAttempts += 1
+ if(!registered) {
+ registrationDuration += registrationDelay
+ // double delay for exponential backoff
+ registrationDelay *= 2
- if (registered) {
- registrationScheduler.foreach(_.cancel())
- }
- else if (registrationAttempts <=
TaskManager.MAX_REGISTRATION_ATTEMPTS) {
+ if (registrationDuration > maxRegistrationDuration) {
+ log.warning("TaskManager could not register at JobManager {}
after {}.", jobManagerAkkaURL,
- log.info("Try to register at master {}. Attempt #{}",
jobManagerAkkaURL,
- registrationAttempts)
- val jobManager = context.actorSelection(jobManagerAkkaURL)
+ maxRegistrationDuration)
- jobManager ! RegisterTaskManager(connectionInfo,
hardwareDescription, numberOfSlots)
- }
- else {
- log.error("TaskManager could not register at JobManager.");
- self ! PoisonPill
+ self ! PoisonPill
+ } else if (!registered) {
+ log.info(s"Try to register at master ${jobManagerAkkaURL}.
${registrationAttempts}. " +
+ s"Attempt")
+ val jobManager = context.actorSelection(jobManagerAkkaURL)
+
+ jobManager ! RegisterTaskManager(connectionInfo,
hardwareDescription, numberOfSlots)
+
+ context.system.scheduler.scheduleOnce(registrationDelay, self,
RegisterAtJobManager)
+ }
}
}
case AcknowledgeRegistration(id, blobPort) => {
- if (!registered) {
+ if(!registered) {
+ finishRegistration(id, blobPort)
registered = true
- currentJobManager = sender
- instanceID = id
-
- context.watch(currentJobManager)
-
- log.info("TaskManager successfully registered at JobManager {}.",
- currentJobManager.path.toString)
-
- setupNetworkEnvironment()
- setupLibraryCacheManager(blobPort)
+ } else {
+ if (log.isDebugEnabled) {
--- End diff --
You're right Henry, I'll remove it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
