You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@sentry.apache.org by Anne Yu <an...@cloudera.com> on 2016/03/23 19:05:20 UTC

Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/
-----------------------------------------------------------

Review request for sentry.


Bugs: SENTRY-1108
    https://issues.apache.org/jira/browse/SENTRY-1108


Repository: sentry


Description
-------

The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.


Diffs
-----

  pom.xml d25c314e0a4c92ea03f22c01c07e0b0beba2bbea 

Diff: https://reviews.apache.org/r/45221/diff/


Testing
-------

Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.

Full suite: 2039 tests, 92m58.727s;
skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;


Thanks,

Anne Yu


Re: Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

Posted by Anne Yu <an...@cloudera.com>.

> On April 9, 2016, 5:53 a.m., Hao Hao wrote:
> > pom.xml, line 900
> > <https://reviews.apache.org/r/45221/diff/2/?file=1321664#file1321664line900>
> >
> >     For a Quad-Core system, if each VM has max heap size to be 2G, and 10 threads will be running simultaneously, so the RAM should be 20G at least? Is it safe to assume that?

Thanks for reviewing this patch.

-Xmx2g is the maximum heap size could be available for the process; but I think it's a valid point. May decrease -Xms512m to -Xms256m; The main test infrastructure I used is pre-commit job machines.


> On April 9, 2016, 5:53 a.m., Hao Hao wrote:
> > pom.xml, line 908
> > <https://reviews.apache.org/r/45221/diff/2/?file=1321664#file1321664line908>
> >
> >     Should we set reuseForks to true, so that to save resource/time for creating a new process for testing every time one test class is done?

Can't. For some tests, if context in one JVM is not cleared up completely, will cause unexpected failures. I've seen failures. But even if reuseForks is false, tests still finish quite fast; reuse or not-reuse doesn't gain much.


> On April 9, 2016, 5:53 a.m., Hao Hao wrote:
> > pom.xml, line 917
> > <https://reviews.apache.org/r/45221/diff/2/?file=1321664#file1321664line917>
> >
> >     Why do we need to exclude TestURI?

It requires dedicated resouces can't share with other JVM processes. As far as why, need to dig more.


- Anne


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/#review127956
-----------------------------------------------------------


On April 1, 2016, 12:03 a.m., Anne Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45221/
> -----------------------------------------------------------
> 
> (Updated April 1, 2016, 12:03 a.m.)
> 
> 
> Review request for sentry and Hao Hao.
> 
> 
> Bugs: SENTRY-1108
>     https://issues.apache.org/jira/browse/SENTRY-1108
> 
> 
> Repository: sentry
> 
> 
> Description
> -------
> 
> The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
> Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.
> 
> 
> Diffs
> -----
> 
>   pom.xml 37db0079503a5a3806526573904a989688a349fe 
> 
> Diff: https://reviews.apache.org/r/45221/diff/
> 
> 
> Testing
> -------
> 
> Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.
> 
> Full suite: 2039 tests, 92m58.727s;
> skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;
> 
> 
> Thanks,
> 
> Anne Yu
> 
>


Re: Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

Posted by Hao Hao <ha...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/#review127956
-----------------------------------------------------------




pom.xml (line 900)
<https://reviews.apache.org/r/45221/#comment191345>

    For a Quad-Core system, if each VM has max heap size to be 2G, and 10 threads will be running simultaneously, so the RAM should be 20G at least? Is it safe to assume that?



pom.xml (line 908)
<https://reviews.apache.org/r/45221/#comment191346>

    Should we set reuseForks to true, so that to save resource/time for creating a new process for testing every time one test class is done?



pom.xml (line 917)
<https://reviews.apache.org/r/45221/#comment191347>

    Why do we need to exclude TestURI?


- Hao Hao


On April 1, 2016, 12:03 a.m., Anne Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45221/
> -----------------------------------------------------------
> 
> (Updated April 1, 2016, 12:03 a.m.)
> 
> 
> Review request for sentry and Hao Hao.
> 
> 
> Bugs: SENTRY-1108
>     https://issues.apache.org/jira/browse/SENTRY-1108
> 
> 
> Repository: sentry
> 
> 
> Description
> -------
> 
> The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
> Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.
> 
> 
> Diffs
> -----
> 
>   pom.xml 37db0079503a5a3806526573904a989688a349fe 
> 
> Diff: https://reviews.apache.org/r/45221/diff/
> 
> 
> Testing
> -------
> 
> Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.
> 
> Full suite: 2039 tests, 92m58.727s;
> skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;
> 
> 
> Thanks,
> 
> Anne Yu
> 
>


Re: Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

Posted by Hao Hao <ha...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/#review128582
-----------------------------------------------------------


Ship it!




Ship It!

- Hao Hao


On April 12, 2016, 11:01 p.m., Anne Yu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/45221/
> -----------------------------------------------------------
> 
> (Updated April 12, 2016, 11:01 p.m.)
> 
> 
> Review request for sentry and Hao Hao.
> 
> 
> Bugs: SENTRY-1108
>     https://issues.apache.org/jira/browse/SENTRY-1108
> 
> 
> Repository: sentry
> 
> 
> Description
> -------
> 
> The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
> Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.
> 
> 
> Diffs
> -----
> 
>   pom.xml c3754ed64d4e361024e4d8910e6d1f2ce86f1be7 
> 
> Diff: https://reviews.apache.org/r/45221/diff/
> 
> 
> Testing
> -------
> 
> Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.
> 
> Full suite: 2039 tests, 92m58.727s;
> skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;
> 
> 
> Thanks,
> 
> Anne Yu
> 
>


Re: Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

Posted by Anne Yu <an...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/
-----------------------------------------------------------

(Updated April 12, 2016, 11:01 p.m.)


Review request for sentry and Hao Hao.


Changes
-------

Addressed code comments regarding xms.


Bugs: SENTRY-1108
    https://issues.apache.org/jira/browse/SENTRY-1108


Repository: sentry


Description
-------

The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.


Diffs (updated)
-----

  pom.xml c3754ed64d4e361024e4d8910e6d1f2ce86f1be7 

Diff: https://reviews.apache.org/r/45221/diff/


Testing
-------

Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.

Full suite: 2039 tests, 92m58.727s;
skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;


Thanks,

Anne Yu


Re: Review Request 45221: Add a test run to exclude slow and not thread safe tests. It could be used for pre-commit jobs

Posted by Anne Yu <an...@cloudera.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/45221/
-----------------------------------------------------------

(Updated April 1, 2016, 12:03 a.m.)


Review request for sentry and Hao Hao.


Changes
-------

Trim more carefully the excluded tests. Totall running time is 15min; However code coverage for core sentry packages is 63.2 (before trim, it is 75%).


Bugs: SENTRY-1108
    https://issues.apache.org/jira/browse/SENTRY-1108


Repository: sentry


Description
-------

The way to run it: mvn test -P skipSlowAndNotThreadSafeTests; So it won't affect default mvn test;
Most excluded slow tests are from sentry-tests/e2e suites and a few are those can't be run under concurrent mode.


Diffs (updated)
-----

  pom.xml 37db0079503a5a3806526573904a989688a349fe 

Diff: https://reviews.apache.org/r/45221/diff/


Testing
-------

Test the running time on an ec2 cluster (large). Run mvn clean install first then mvn test.

Full suite: 2039 tests, 92m58.727s;
skipSlowAndNotThreadSafeTests run: 1027 tests, 6m47.195s;


Thanks,

Anne Yu