[jira] [Commented] (FLINK-1481) Flakey JobManagerITCase

["ASF GitHub Bot (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-1481?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14307321#comment-14307321 ] 

ASF GitHub Bot commented on FLINK-1481:
---------------------------------------

GitHub user tillrohrmann opened a pull request:

    https://github.com/apache/flink/pull/365

    [FLINK-1481] Fixes flakey JobManagerITCase

    The sometimes failing sender tasks are now guaranteed to have at least one task which fails.

You can merge this pull request into a Git repository by running:

    $ git pull https://github.com/tillrohrmann/flink fixFlakeyJobManagerITCase

Alternatively you can review and apply these changes as the patch at:

    https://github.com/apache/flink/pull/365.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 #365
    
----
commit 0964a6ebaa64c9ae644831564ff898f23e851bb6
Author: Till Rohrmann <tr...@apache.org>
Date:   2015-02-05T14:48:23Z

    [FLINK-1481] Fixes flakey JobManagerITCase which relied on non-deterministic behaviour.

----


> Flakey JobManagerITCase
> -----------------------
>
>                 Key: FLINK-1481
>                 URL: https://issues.apache.org/jira/browse/FLINK-1481
>             Project: Flink
>          Issue Type: Bug
>            Reporter: Till Rohrmann
>            Priority: Minor
>
> We currently have some test cases which rely on non-deterministic behaviour. For example the {{JobManagerItCase}} contains a test case with a {{SometimesExceptionSender}} which randomly throws an exception and otherwise blocks. The probability of a failure is 0.05 and we have 100 parallel tasks. Thus, the overall probability that no task fails at all is still 0.005. Thus every 200th test run, the test case will block and thus fail.
> In order to get a stable test case base, I think that we should try to avoid these kind of test cases with random behaviour. The same goes with sleep timeouts in order to "establish" an intended interleaving of concurrent processes. With Travis this can fail too often.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)