[
https://issues.apache.org/jira/browse/IGNITE-10225?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16712910#comment-16712910
]
ASF GitHub Bot commented on IGNITE-10225:
-----------------------------------------
GitHub user DirectXceriD opened a pull request:
https://github.com/apache/ignite/pull/5606
IGNITE-10225 fix starvation check
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/gridgain/apache-ignite ignite-10225
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/ignite/pull/5606.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 #5606
----
commit ec869715a4d3521715ac08d1112ba5e61e325098
Author: Vladimir Pligin <vova199107@...>
Date: 2018-12-07T14:07:45Z
IGNITE-10225 add new system property
----
> Striped pool starvation check reports false-positive warnings if a job starts
> right before the check
> ----------------------------------------------------------------------------------------------------
>
> Key: IGNITE-10225
> URL: https://issues.apache.org/jira/browse/IGNITE-10225
> Project: Ignite
> Issue Type: Bug
> Reporter: Stanislav Lukyanov
> Assignee: Vladimir Pligin
> Priority: Major
> Labels: newbie
> Attachments: StripedExecutorTest.java
>
>
> StripedExecutor::checkStarvation checks if two conditions are met: a stripe
> is executing a job and the stripe's job counter hasn't changed since the last
> check.
> These conditions will be met if no jobs have been executed for a long time
> and a job (even a very short one) has just started.
> This check isn't supposed to be exact and allowed to produce false-positives.
> However, the check is supposed to detect jobs running for more than
> starvation check interval (IGNITE_STARVATION_CHECK_INTERVAL), but in this
> case it may detect even very short ones.
> To fix this we may require a stripe to have no progress in two consecutive
> checks - then we will have much less chance for a false-positive, but still
> detect hanging or very long jobs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
