One culprit will be the Ranger authorization tests in AuthorizationStmtTest since we now run the tests against both Sentry and Ranger: https://github.com/apache/impala/blob/931a8f0ba7f45d5b1608e62aff397b517b943e95/fe/src/test/java/org/apache/impala/analysis/AuthorizationStmtTest.java#L171-L174. However, this gives us a good coverage and also ensures similar behavior between Sentry and Ranger. I hope this is a good trade off.
On Mon, Apr 29, 2019 at 10:32 AM Tim Armstrong <tarmstr...@cloudera.com> wrote: > It looks like 30 minutes is somehow lost in frontend tests: > > > *02:20:56* [INFO] Tests run: 10, Failures: 0, Errors: 0, Skipped: 0, > Time elapsed: 7.156 s - in > org.apache.impala.authorization.sentry.SentryProxyTest*02:20:56* > [INFO] Running org.apache.impala.catalog.CatalogTest*02:52:05* [INFO] > Tests run: 10, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 3.749 > s - in org.apache.impala.catalog.CatalogTest*02:52:05* [INFO] Running > org.apache.impala.catalog.CatalogObjectToFromThriftTest*02:52:05* > [INFO] Tests run: 6, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: > 0.736 s - in > org.apache.impala.catalog.CatalogObjectToFromThriftTest*02:52:05* > [INFO] Running > org.apache.impala.catalog.events.MetastoreEventsProcessorTest*02:52:05* > [INFO] Tests run: 30, Failures: 0, Errors: 0, Skipped: 0, Time > elapsed: 187.46 s - in > org.apache.impala.catalog.events.MetastoreEventsProcessorTest > > > I'm not sure where the other increase is coming from - may just be organic > growth in the number of tests and we need to cut down runtime elsewhere. > > On Sun, Apr 28, 2019 at 11:17 AM Jim Apple <jbap...@apache.org> wrote: > > > The major from-scratch job in the pre-merge tests seems to frequently be > > taking more than 5 hours. It used to frequently take less than 4, as of a > > month or two ago. > > > > Here's a job that calls that, to show how long the job has been taking > > recently: > > > > > https://jenkins.impala.io/view/Utility/job/parallel-all-tests/buildTimeTrend > > > > Was something big added that increased the run time? It's possible I > missed > > an email thread about this. > > >