cloud-fan commented on a change in pull request #29817:
URL: https://github.com/apache/spark/pull/29817#discussion_r509858066
##########
File path:
core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala
##########
@@ -213,11 +204,31 @@ private[spark] class CoarseGrainedExecutorBackend(
logInfo(s"Received tokens of ${tokenBytes.length} bytes")
SparkHadoopUtil.get.addDelegationTokens(tokenBytes, env.conf)
- case DecommissionSelf =>
- logInfo("Received decommission self")
+ case DecommissionExecutor =>
decommissionSelf()
}
+ override def receiveAndReply(context: RpcCallContext): PartialFunction[Any,
Unit] = {
+ case ExecutorSigPWRReceived =>
+ var driverNotified = false
+ try {
+ driver.foreach { driverRef =>
+ // Tell driver that we are starting decommissioning so it stops
trying to schedule us
+ driverNotified =
driverRef.askSync[Boolean](ExecutorDecommissioning(executorId))
+ if (driverNotified) decommissionSelf()
Review comment:
This part is new. I think it makes sense to skip executor
decommissioning if the driver side rejects to do so.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
