[
https://issues.apache.org/jira/browse/SPARK-20843?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16026663#comment-16026663
]
Michael Allman commented on SPARK-20843:
----------------------------------------
bq. I will say that I don't think its really safe to rely on clean shutdowns
for correctness...
I agree, but we don't want a "force kill" to become the norm for an app
shutdown. An unclean shutdown should be an exceptional situation that will be
cause for an ops alert, an investigation and a validation that no data loss or
corruption occurred (i.e. our failure mechanisms held).
Also, I would say that in some cases where an app integrates with other systems
and cannot guarantee transactional or idempotent semantics in the event of a
failure, an unclean shutdown *will* require cross-system validation and any
necessary data synchronization or recovery.
Cheers.
> Cannot gracefully kill drivers which take longer than 10 seconds to die
> -----------------------------------------------------------------------
>
> Key: SPARK-20843
> URL: https://issues.apache.org/jira/browse/SPARK-20843
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 2.1.1
> Reporter: Michael Allman
> Labels: regression
>
> Commit
> https://github.com/apache/spark/commit/1c9a386c6b6812a3931f3fb0004249894a01f657
> changed the behavior of driver process termination. Whereas before
> `Process.destroyForcibly` was never called, now it is called (on Java VM's
> supporting that API) if the driver process does not die within 10 seconds.
> This prevents apps which take longer than 10 seconds to shutdown gracefully
> from shutting down gracefully. For example, streaming apps with a large batch
> duration (say, 30 seconds+) can take minutes to shutdown.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
