Ngone51 commented on a change in pull request #29722:
URL: https://github.com/apache/spark/pull/29722#discussion_r487937373
##########
File path:
core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala
##########
@@ -79,12 +79,23 @@ private[spark] class CoarseGrainedExecutorBackend(
*/
private[executor] val taskResources = new mutable.HashMap[Long, Map[String,
ResourceInformation]]
- @volatile private var decommissioned = false
+ private var decommissioned = false
override def onStart(): Unit = {
- logInfo("Registering PWR handler.")
- SignalUtils.register("PWR", "Failed to register SIGPWR handler - " +
- "disabling decommission feature.")(decommissionSelf)
+ if (env.conf.get(DECOMMISSION_ENABLED)) {
+ logInfo("Registering PWR handler to trigger decommissioning.")
+ SignalUtils.register("PWR", "Failed to register SIGPWR handler - " +
+ "disabling executor decommission feature.") {
+ self.send(DecommissionExecutor)
Review comment:
@cloud-fan @agrawaldevesh What do you think about the latest change? In
case of executor is decommissioned by the signal, we send
`DecommissionExecutor`(it's a local RPC call) to the executor itself instead of
invoking `decommissionSelf` directly.
In this way, we can guarantee the thread-safe of `decommissioned`.
(It's same to the Worker.)
----------------------------------------------------------------
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]
