> On Jan. 25, 2018, 4:16 p.m., Alexander Kolbasov wrote: > > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/provider/db/service/persistent/TransactionManager.java > > Line 231 (original), 246 (patched) > > <https://reviews.apache.org/r/65268/diff/3/?file=1944703#file1944703line246> > > > > If you don't think that you can use time because we do not know how > > much time it may take to execute transaction, then the whole point of this > > JIRA is moot and we shouldn't fix it in the first place. > > Vadim Spector wrote: > I am not sure if it's the real issue. The two limita are not redundant or > mutually exclusive. The old approach limits the number of re-tries; this > change also adds limits the total re-try time but keeps the old limitation as > well. Are two controls necessarily worse than one? If someone re-configures > by mistake retry interval too small (much smaller than the default), so there > may be lots of retries (and possibly very unresponsive system) before time > limit reached - perhaps we want to prevent it? Can system clock reset at run > time messing with the time-only logic? Unlikely, but not impossible.
Vadim, Having both limits on retry-count and the retry-time would address the scenario you are taking about. - kalyan kumar ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/65268/#review196263 ----------------------------------------------------------- On Jan. 23, 2018, 11:40 p.m., kalyan kumar kalvagadda wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/65268/ > ----------------------------------------------------------- > > (Updated Jan. 23, 2018, 11:40 p.m.) > > > Review request for sentry, Arjun Mishra, Na Li, Sergio Pena, and Vadim > Spector. > > > Bugs: SENTRY-1904 > https://issues.apache.org/jira/browse/SENTRY-1904 > > > Repository: sentry > > > Description > ------- > > The TransactionManager uses exponential backoff strategy for transaction > retries. This may cause some transactions to be delayed by a very long time. > We should also have a constraint on the max time for a transaction so that we > do not retry for too long. > > New patch that is attached adds upper bounds on below > > 1.Interval between the retry attempts which increases exponentially. > 2.Total time a transaction could spend in retries. > > > With out these limits we would not have a control on how long a transaction > could be be active. > > > Diffs > ----- > > > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/provider/db/service/persistent/HMSFollower.java > 2f2b98412e7dfdcc847ffe7975a70f452554e747 > > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/provider/db/service/persistent/SentryStore.java > edea5b64d8f98c93aafc1fe43fa97e00c2ce2948 > > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/provider/db/service/persistent/TransactionManager.java > f4ff962a67f8a5c23cc5c8daa7bcb861d2e6b6a5 > > sentry-provider/sentry-provider-db/src/main/java/org/apache/sentry/service/thrift/ServiceConstants.java > 7e02874b4be6a7109108809b1f404fe971b7b8e2 > > sentry-provider/sentry-provider-db/src/test/java/org/apache/sentry/provider/db/service/persistent/TestSentryStore.java > b4100278392986c161625a366212c6fef66ec0a9 > > > Diff: https://reviews.apache.org/r/65268/diff/3/ > > > Testing > ------- > > Made sure all the tests pass. > > > Thanks, > > kalyan kumar kalvagadda > >
