[
https://issues.apache.org/jira/browse/HADOOP-6810?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12877246#action_12877246
]
Jakob Homan commented on HADOOP-6810:
-------------------------------------
I tend to agree with Boris. With the current patch, if I run test-core, the
behavior changes so that those tests in commit and smoke will *not* be run,
correct? That would be a big change from what users have grown used to. I had
assumed that commit tests are a subset of smoke tests are a subset of
core-tests are a subset of test. I had made previous comments that if you run,
say ant commit-tests test, you would end up running the members of commit-tests
twice, which was suboptimal. But I'm not sure the other side of that - three
separate disjoint sets - is also optimal.
Except for this behavior change, the patch looks good.
> Extract a subset of tests for smoke (DOA) validation
> ----------------------------------------------------
>
> Key: HADOOP-6810
> URL: https://issues.apache.org/jira/browse/HADOOP-6810
> Project: Hadoop Common
> Issue Type: Improvement
> Components: test
> Affects Versions: 0.21.0
> Reporter: Konstantin Boudnik
> Assignee: Konstantin Boudnik
> Attachments: smoke-tests.patch, smoke-tests.patch, smoke-tests.patch,
> smoke-tests.patch
>
>
> Currently, there's no way to quickly find out if a new build of Hadoop is
> reasonably viable and can be used for future development/testing cycle.
> Similarly to 'commit' testing it'd be great to have a wider subset of
> existing tests to be executed inside of 30 minutes. The main goal of such a
> subset is to have a reasonable guarantees that different parts of Hadoop's
> functionality aren't broken and the system is not DOA.
> Full test run takes about 3+ hours and it is a too long wait time in many
> cases.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
