[
https://issues.apache.org/jira/browse/FLINK-3107?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15139047#comment-15139047
]
ASF GitHub Bot commented on FLINK-3107:
---------------------------------------
GitHub user uce opened a pull request:
https://github.com/apache/flink/pull/1610
[FLINK-3107] [runtime] Start checkpoint ID counter with periodic scheduler
Problem: The job manager enables checkpoints during submission of streaming
programs. This can lead to call to a call to
`ZooKeeperCheckpointIDCounter.start()`, which communicates with ZooKeeper. This
can block the job manager actor.
Solution: Start the counter in the `CheckpointCoordinatorDeActivator`.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/uce/flink 3107-counter_start
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/1610.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 #1610
----
commit d70bc79e48dddb658c2240350837000ce9f1f0fe
Author: Ufuk Celebi <[email protected]>
Date: 2016-02-09T15:06:46Z
[FLINK-3107] [runtime] Start checkpoint ID counter with periodic scheduler
Problem: The job manager enables checkpoints during submission of streaming
programs. This can lead to call to a call to
`ZooKeeperCheckpointIDCounter.start()`,
which communicates with ZooKeeper. This can block the job manager actor.
Solution: Start the counter in the `CheckpointCoordinatorDeActivator`.
----
> ZooKeeperCheckpointIDCounter.start() can block JobManager actor
> ---------------------------------------------------------------
>
> Key: FLINK-3107
> URL: https://issues.apache.org/jira/browse/FLINK-3107
> Project: Flink
> Issue Type: Bug
> Components: JobManager
> Affects Versions: 0.10.1
> Reporter: Ufuk Celebi
>
> In HA mode, the job manager enables checkpoints during submission of
> streaming programs.
> This leads to call to ZooKeeperCheckpointIDCounter.start(), which
> communicates with ZooKeeper. This can block the job manager actor.
> A solution is to start the counter later instead of the CheckpointCoordinator
> constructor.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
