[
https://issues.apache.org/jira/browse/BEAM-1674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15904835#comment-15904835
]
ASF GitHub Bot commented on BEAM-1674:
--------------------------------------
GitHub user aljoscha opened a pull request:
https://github.com/apache/beam/pull/2217
[BEAM-1674] Fix Flink State GC
This is a proper solution, as discussed in the Jira issue. If we merge this
we can drop #2215. (Thanks for quickly providing that PR, though!)
R: @kennknowles
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/aljoscha/beam jira-1674-fix-flink-gc
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/beam/pull/2217.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #2217
----
commit bf6d2748c8876a7415290069163625598928f02f
Author: Aljoscha Krettek <aljoscha.kret...@gmail.com>
Date: 2017-03-10T07:29:27Z
Move GC timer checking to StatefulDoFnRunner.CleanupTimer
commit 1a8e1f7463cbc7c6b5edfe1dbbc98502e5612511
Author: Aljoscha Krettek <aljoscha.kret...@gmail.com>
Date: 2017-03-10T10:07:00Z
Introduce Flink-specific state GC implementations
We now set the GC timer for window.maxTimestamp() + 1 to ensure that a
user timer set for window.maxTimestamp() still has all state.
This also adds tests for late data dropping and state GC specifically
for the Flink DoFnOperator.
----
> Flink user state GC depends on order of timer firing
> ----------------------------------------------------
>
> Key: BEAM-1674
> URL: https://issues.apache.org/jira/browse/BEAM-1674
> Project: Beam
> Issue Type: Bug
> Components: runner-flink
> Affects Versions: 0.5.0
> Reporter: Aljoscha Krettek
> Assignee: Aljoscha Krettek
> Priority: Blocker
> Fix For: 0.6.0
>
>
> The newly added {{ParDoTest.testEventTimeTimerMultipleKeys()}} fails because
> the {{DoFn}} sets a timer for {{window.maxTimestamp()}} which also happens to
> be the GC timer for the user state. The Flink Runner uses timers to schedule
> GC, the user-set timer and the GC timer have a different timer id, so they
> don't clash. However, if the GC timer is being processed before the user
> timer then the user doesn't have a chance to access the state anymore because
> it will already be cleared out by the time the user timer is being processed.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
