[jira] [Commented] (HADOOP-11617) Several tests are not stable (on OpenJDK / x86_64 / RHEL 7.1)

["Tony Reix (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HADOOP-11617?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14345099#comment-14345099 ] 

Tony Reix commented on HADOOP-11617:
------------------------------------

Hi Steve, I cannot do the same on the hadoop-trunk. I have not enough machines, this stability test lasts about 11 days, and I have to focus on Hadoop 2.6.0 now, on PPC64LE.
However, as I said, I'm doing the same work with Hadoop 2.6.0. See: https://issues.apache.org/jira/browse/HADOOP-11636 .
However, your recommendation to search for tests names is good advice. I'll do.
Moreover, several of these unstable tests with OpenJDK are nearly stable (and fail) with IBM JVM. So, I would recommend to use both OpenJDK and IBM JVM for testing Hadoop. That would speed up finding complex & hidden issues.

> Several tests are not stable (on OpenJDK / x86_64 / RHEL 7.1)
> -------------------------------------------------------------
>
>                 Key: HADOOP-11617
>                 URL: https://issues.apache.org/jira/browse/HADOOP-11617
>             Project: Hadoop Common
>          Issue Type: Bug
>         Environment: RHEL 7.1 / x86_64
>            Reporter: Tony Reix
>
> I'm now porting Hadoop 2.4.1 on PPC64 HW.
> In order to have an idea about the stability of the tests, I've run several times all Hadoop tests in a RHEL 7.1 / x86_64 / OpenJDK environment.
> I've run it 12 times, for now, on the EXACT same environment.
> Using some tool, I have extracted the list of tests that do not always produce the same results, listed below, showing some test unstability.
> I 'm now doing the same for version 2.6.0 on OpenJDK/Ubuntu/x86_64.
> Report gives:
> test-name N: T F E S |
> where:
> - N is the number of the test log file
> - and:
> T: Tests run
> F: Failure
> E: Error
> S: Skipped
> As an example (see below for full data):
> hdfs.server.datanode.TestDeleteBlockPool                      0:    2   0   0   0 |
> ........
> hdfs.server.datanode.TestDeleteBlockPool                      4:    2   0   1   0 |
> ........
> hdfs.server.datanode.TestDeleteBlockPool                     11:    2   0   0   0 |
> means that this test, out of 12 runs, had an error once, with 0 Failure and 0 Skipped test.



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