[
https://issues.apache.org/jira/browse/BEAM-5865?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16722919#comment-16722919
]
Jozef Vilcek commented on BEAM-5865:
------------------------------------
I do not strictly want to write partitions as they are. I want to write kafka
data effectively. To be able to do it without shuffling, would be great.
However I am under impression that windowing requires keyed pcollection which
will result in shuffle.
As a first step, I am even OK with shuffle, but do not want to end up with hot
spots after shuffle where one worker is managing e.g. 3 keys -> writing 3 files
and holding that state before flush and some other workers doing zero keys.
How can I do custom transform + make sure that under errors / restarts I will
not be observing duplicate events in this transform? Can not see a way how to
make this work with state. I will be grateful for any hints.
> 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)
