[
https://issues.apache.org/jira/browse/BEAM-5865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16703642#comment-16703642
]
Maximilian Michels commented on BEAM-5865:
------------------------------------------
I think the stable input is relevant for continuing to write from a checkpoint,
especially when exactly-once semantics should be obeyed which requires
idempotent writes. For files this should not be possible anyways.
The key is assigned randomly as of now. It's in the
{{WritesFiles$WriteShardedBundlesToTempFiles}}, but we could decide to use a
different partitioning strategy for achieving balanced partitions. But we can
only do that if we know that we're not executing a normal Reshuffle. So the
code there would have to be altered to use a custom URN different from
Reshuffle.
We would then have to embed translation code in the FlinkRunner translation
classes where we can influence the partitioning. A simple transform replacement
would not be enough, as you already pointed out.
> Auto sharding of streaming sinks in FlinkRunner
> -----------------------------------------------
>
> Key: BEAM-5865
> URL: https://issues.apache.org/jira/browse/BEAM-5865
> Project: Beam
> Issue Type: Improvement
> Components: runner-flink
> Reporter: Maximilian Michels
> Priority: Major
>
> The Flink Runner should do auto-sharding of streaming sinks, similar to
> BEAM-1438. That way, the user doesn't have to set shards manually which
> introduces additional shuffling and might cause skew in the distribution of
> data.
> As per discussion:
> https://lists.apache.org/thread.html/7b92145dd9ae68da1866f1047445479f51d31f103d6407316bb4114c@%3Cuser.beam.apache.org%3E
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
