[
https://issues.apache.org/jira/browse/SPARK-11141?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15902537#comment-15902537
]
Jim Kleckner edited comment on SPARK-11141 at 3/9/17 5:54 AM:
--------------------------------------------------------------
FYI, this can cause problems when not using S3 during shutdown as described in
this AWS posting: https://forums.aws.amazon.com/thread.jspa?threadID=223378
The workaround indicated is to use --conf
spark.streaming.driver.writeAheadLog.allowBatching=false with the submit.
The exception contains the text:
{code}
streaming stop ReceivedBlockTracker: Exception thrown while writing record:
BatchAllocationEvent
{code}
was (Author: jkleckner):
FYI, this can cause problems when not using S3 during shutdown as described in
this AWS posting: https://forums.aws.amazon.com/thread.jspa?threadID=223378
The workaround indicated is to use --conf
spark.streaming.driver.writeAheadLog.allowBatching=false with the submit.
> Batching of ReceivedBlockTrackerLogEvents for efficient WAL writes
> ------------------------------------------------------------------
>
> Key: SPARK-11141
> URL: https://issues.apache.org/jira/browse/SPARK-11141
> Project: Spark
> Issue Type: Improvement
> Components: DStreams
> Reporter: Burak Yavuz
> Assignee: Burak Yavuz
> Fix For: 1.6.0
>
>
> When using S3 as a directory for WALs, the writes take too long. The driver
> gets very easily bottlenecked when multiple receivers send AddBlock events to
> the ReceiverTracker. This PR adds batching of events in the
> ReceivedBlockTracker so that receivers don't get blocked by the driver for
> too long.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
