> On May 31, 2016, at 8:53 PM, Andrew Wang <andrew.w...@cloudera.com> wrote:
>
> I'm in favor of something that helps unify the current mess of Jenkins
> jobs. We do something similar for our internal Hadoop repo: each branch has
> a "build.sh" and "test.sh" script that builds and then runs the tests. This
> predates Yetus, else we'd probably have used that. So +1 from me.
>
> One thing I also find really helpful is riding over flaky tests, perhaps
> via:
>
> http://maven.apache.org/surefire/maven-surefire-plugin/examples/rerun-failing-tests.html
>
> This would be great for precommit too, since right now someone has to run
> the failed tests locally to confirm they're flakes.
>
> Even better is then feeding the success/failure into a DB so you can track
> flaky rates over time, which helps you prioritize which flaky to fix first,
> and helps you find the change which caused the flakiness.
Appy over in HBase-land has added flaky detection and the ability to
track and ignore them to their Yetus personality. At some point, I’m sure
we’ll see that code float upstream in a more generalized manner as they get
more experience with it.
---------------------------------------------------------------------
To unsubscribe, e-mail: common-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-dev-h...@hadoop.apache.org
