[
https://issues.apache.org/jira/browse/BEAM-9979?focusedWorklogId=472959&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-472959
]
ASF GitHub Bot logged work on BEAM-9979:
----------------------------------------
Author: ASF GitHub Bot
Created on: 20/Aug/20 16:14
Start Date: 20/Aug/20 16:14
Worklog Time Spent: 10m
Work Description: lukecwik commented on a change in pull request #12639:
URL: https://github.com/apache/beam/pull/12639#discussion_r474104405
##########
File path:
sdks/java/harness/src/main/java/org/apache/beam/fn/harness/BeamFnDataReadRunner.java
##########
@@ -237,8 +232,9 @@ public void trySplit(
}
synchronized (splittingLock) {
- // Don't attempt to split if we haven't started.
- if (!started) {
+ // Don't attempt to split if we are already done since there isn't a
meaningful split we can
+ // provide.
+ if (index == stopIndex) {
Review comment:
No, the logic below will allow us to choose a stopIndex even if
`registerInputLocation` has not been invoked. See
`testSplittingWhenNoElementsProcessed`. I added a variant of that test where we
split before `registerInputLocation` happens to get coverage for this case.
----------------------------------------------------------------
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: 472959)
Time Spent: 1h 10m (was: 1h)
> Fix race condition where the read index maybe reported from the last executed
> bundle
> ------------------------------------------------------------------------------------
>
> Key: BEAM-9979
> URL: https://issues.apache.org/jira/browse/BEAM-9979
> Project: Beam
> Issue Type: Bug
> Components: sdk-java-harness
> Affects Versions: 2.21.0, 2.22.0, 2.23.0, 2.24.0
> Reporter: Luke Cwik
> Assignee: Luke Cwik
> Priority: P3
> Fix For: 2.25.0
>
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> When the BeamFnDataReadRunner is reused there is a short period of time when
> a progress request could happen before the the start function is called
> resetting the read index to -1.
> I believe there should be a way to *reset* an operator before it gets added
> to the set of cached bundle processors separate instead of placing clean-up
> in any *start* functions that those operators may rely on preventing exposing
> details of those operators before *start* may have been invoked.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
