[
https://issues.apache.org/jira/browse/BEAM-9900?focusedWorklogId=435363&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-435363
]
ASF GitHub Bot logged work on BEAM-9900:
----------------------------------------
Author: ASF GitHub Bot
Created on: 20/May/20 06:40
Start Date: 20/May/20 06:40
Worklog Time Spent: 10m
Work Description: dmvk merged pull request #11750:
URL: https://github.com/apache/beam/pull/11750
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 435363)
Time Spent: 1.5h (was: 1h 20m)
> Remove the need for shutdownSourcesOnFinalWatermark flag
> --------------------------------------------------------
>
> Key: BEAM-9900
> URL: https://issues.apache.org/jira/browse/BEAM-9900
> Project: Beam
> Issue Type: Improvement
> Components: runner-flink
> Reporter: Maximilian Michels
> Assignee: Maximilian Michels
> Priority: P2
> Fix For: 2.22.0
>
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> The {{shutdownSourcesOnFinalWatermark}} has caused some confusion in the
> past. It is generally used for testing pipelines to ensure that the pipeline
> and the testing cluster shuts down at the end of the job. Without it, the
> pipeline will run forever in streaming mode, regardless of whether the input
> is finite or not.
> We didn't want to enable the flag by default because shutting down any
> operators including sources in Flink will prevent checkpointing from working.
> If we have side input, for example, that may be the case even for
> long-running pipelines. However, we can detect whether checkpointing is
> enabled and set the flag automatically.
> The only situation where we may want the flag to be disabled is when users do
> not have checkpointing enabled but want to take a savepoint. This should be
> rare and users can mitigate by setting the flag to false to prevent operators
> from shutting down.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
