[
https://issues.apache.org/jira/browse/BEAM-1438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16951897#comment-16951897
]
Alexey Romanenko commented on BEAM-1438:
----------------------------------------
Does it actually work with Dataflow for file-based IOs? As I see, every IO that
uses {{WriteFiles.wuthNumShards(0)}} will fail for unbounded source because of
the check hat Robert mentioned above. Am I mistaken?
> The default behavior for the Write transform doesn't work well with the
> Dataflow streaming runner
> -------------------------------------------------------------------------------------------------
>
> Key: BEAM-1438
> URL: https://issues.apache.org/jira/browse/BEAM-1438
> Project: Beam
> Issue Type: Bug
> Components: runner-dataflow
> Reporter: Reuven Lax
> Assignee: Reuven Lax
> Priority: Major
> Fix For: 2.5.0
>
>
> If a Write specifies 0 output shards, that implies the runner should pick an
> appropriate sharding. The default behavior is to write one shard per input
> bundle. This works well with the Dataflow batch runner, but not with the
> streaming runner which produces large numbers of small bundles.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
