[
https://issues.apache.org/jira/browse/BEAM-12475?focusedWorklogId=610979&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-610979
]
ASF GitHub Bot logged work on BEAM-12475:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 14/Jun/21 22:25
Start Date: 14/Jun/21 22:25
Worklog Time Spent: 10m
Work Description: robertwb commented on a change in pull request #14991:
URL: https://github.com/apache/beam/pull/14991#discussion_r651317762
##########
File path:
sdks/java/harness/src/test/java/org/apache/beam/fn/harness/BeamFnDataReadRunnerTest.java
##########
@@ -178,7 +179,7 @@ public void testCreatingAndProcessingBeamFnDataReadRunner()
throws Exception {
null /* beamFnTimerClient */,
Review comment:
https://github.com/apache/beam/pull/15011
I would not be oppose to adding additional locking to ProcessBundleHandler,
but in general we try to avoid locks while invoking user code so this seems to
imply we still need to have locking in BeamFnDataReadRunner (as well as it
being good practice to guard against internal access of private variables that
may be accessed from multiple threads).
--
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: 610979)
Time Spent: 1h 40m (was: 1.5h)
> When bundle processors are re-used, do not respond to splits for previous
> bundles.
> ----------------------------------------------------------------------------------
>
> Key: BEAM-12475
> URL: https://issues.apache.org/jira/browse/BEAM-12475
> Project: Beam
> Issue Type: Bug
> Components: java-fn-execution
> Affects Versions: 2.30.0
> Reporter: Robert Bradshaw
> Assignee: Robert Bradshaw
> Priority: P1
> Time Spent: 1h 40m
> Remaining Estimate: 0h
>
> The error manifests in the following way:
> (thread1) Start processing bundle A
> (thread2) Receive request to split for bundle A
> (thread1) Finish bundle A
> (thread1) Reset bundle processor
> (thread1) Start processing bundle B
> (thread2) Process split for bundle A on bundle B
> Runners, such as Dataflow, that ensure the self-reported number of elements
> processed by the SDK is equal to the number they expected to be processed
> (taking into account the runner's understanding of all splits) can detect
> this and reject A and B as invalid, but we should fix this race condition.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
