You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@kafka.apache.org by "ASF GitHub Bot (JIRA)" <ji...@apache.org> on 2016/07/17 16:32:20 UTC
[jira] [Commented] (KAFKA-3812) State store locking is incorrect
[ https://issues.apache.org/jira/browse/KAFKA-3812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15381412#comment-15381412 ]
ASF GitHub Bot commented on KAFKA-3812:
---------------------------------------
GitHub user dguy opened a pull request:
https://github.com/apache/kafka/pull/1628
KAFKA-3812 State store locking is incorrect
Move all state directory creation/locking/unlocking/cleaning to a single class. Don't release the channel until the lock is released. Refactor code to make use of new class
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/dguy/kafka kafka-3812
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/1628.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 #1628
----
commit b4547162d5d07cd471c1ccf9e84ed57c947bf2e4
Author: Damian Guy <da...@gmail.com>
Date: 2016-07-17T16:29:20Z
move locking/cleaning/creating/unlocking of task directories into a single class. refactor code to make use of it
----
> State store locking is incorrect
> --------------------------------
>
> Key: KAFKA-3812
> URL: https://issues.apache.org/jira/browse/KAFKA-3812
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Affects Versions: 0.10.1.0
> Reporter: Eno Thereska
> Assignee: Damian Guy
> Fix For: 0.10.1.0
>
>
> In {{ProcessorStateManager}} in the {{lockStateDirectory}} method, we close a channel if the lock is null. However, as described in https://docs.oracle.com/javase/7/docs/api/java/nio/channels/FileLock.html, "...closing a channel releases all locks held by the Java virtual machine on the underlying file regardless of whether the locks were acquired via that channel or via another channel open on the same file. It is strongly recommended that, within a program, a unique channel be used to acquire all locks on any given file.". Thus closing this channel sometimes leads to the other lock in {{ProcessorStateManager}}, called {{directoryLock}} to be released.
> Problem is reproduced as part of the steps in KAFKA-3805, when two processes are using the same state directory.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)