> On Jan. 25, 2019, 9:01 p.m., kalyan kumar kalvagadda wrote: > > sentry-tests/sentry-tests-hive/src/test/java/org/apache/sentry/tests/e2e/minisentry/InternalSentrySrv.java > > Lines 150 (patched) > > <https://reviews.apache.org/r/69840/diff/1/?file=2122355#file2122355line150> > > > > You are adding 250 msec delay. Is it good enough? > > > > There are few tests that have sentryServers.size()> 1, so we should be > > good even with higher delay. > > > > > > Based on your tests if you think that this delay is good enough, i'm > > fine. > > Na Li wrote: > Higher delay is not necessarily good to avoid deadlock. The table > creation happens only when HMS followers gets notification and create table, > which happens after all sentry services started. So starting services with > delay itself does not fix the issue. What really fixes the issue is the space > out when HMS follower threads run. Those threads run periodically. The > biggest space between two threads is half of the cycle. The biggest space for > N threads is cycle/N. > > I can make the code change to handle thread number > 2.
I have changed the code, so the first instance of sentry server does not wait (this avoid unnecessary wait in test). Only the following instances will wait for certain time to avoid creating table at the same time. - Na ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/69840/#review212353 ----------------------------------------------------------- On Jan. 28, 2019, 4:19 p.m., Na Li wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/69840/ > ----------------------------------------------------------- > > (Updated Jan. 28, 2019, 4:19 p.m.) > > > Review request for sentry, Arjun Mishra, Haley Reeve, and kalyan kumar > kalvagadda. > > > Bugs: sentry-2491 > https://issues.apache.org/jira/browse/sentry-2491 > > > Repository: sentry > > > Description > ------- > > In sentry unit tests, we don't create schema before running a test. Instead, > we use dataNucleus to create sentry tables when they are accessed. This > creates potential deadlock when running test for Sentry HA setup. > > The solution is to let the instances of sentry service start with delay. > Specifically, > let HMS follower threads separate as far as possible, i.e., half of the > interval. > > This deadlock only exists in unit tests, and does not exist in production > because schema is created before starting Sentry services. Therefore, there > is no table creation after service starts. > > > Diffs > ----- > > > sentry-tests/sentry-tests-hive/src/test/java/org/apache/sentry/tests/e2e/minisentry/InternalSentrySrv.java > 9fa42f2 > > > Diff: https://reviews.apache.org/r/69840/diff/3/ > > > Testing > ------- > > such deadlock does not happen with this fix. > Other unit tests passed > > > Thanks, > > Na Li > >
