[
https://issues.apache.org/jira/browse/SPARK-25449?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16791960#comment-16791960
]
Shixiong Zhu commented on SPARK-25449:
--------------------------------------
I think this patch actually fixed a bug introduced by
https://github.com/apache/spark/commit/0514e8d4b69615ba8918649e7e3c46b5713b6540
It didn't use the correct default timeout. Before this batch, using
`spark.executor.heartbeatInterval 30` would send a heartbeat every 30 ms, but
each heartbeat RPC message timeout was 30 seconds.
This patch just unifies the default time unit in all usages of
"spark.executor.heartbeatInterval".
> Don't send zero accumulators in heartbeats
> ------------------------------------------
>
> Key: SPARK-25449
> URL: https://issues.apache.org/jira/browse/SPARK-25449
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 2.4.0
> Reporter: Mukul Murthy
> Assignee: Mukul Murthy
> Priority: Major
> Labels: release-notes
> Fix For: 3.0.0
>
>
> Heartbeats sent from executors to the driver every 10 seconds contain metrics
> and are generally on the order of a few KBs. However, for large jobs with
> lots of tasks, heartbeats can be on the order of tens of MBs, causing tasks
> to die with heartbeat failures. We can mitigate this by not sending zero
> metrics to the driver.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
