https://github.com/cloudera/kudu/commit/679c6b2c93313bd4371916678b4ea8f52b621a50 is desirable.
https://github.com/cloudera/kudu/commit/5f7d79910d8da492195b5c22a6d9a7911f950107 doesn't matter as much, because Shutdown() is never actually invoked in production (people just kill their masters). But, if you want to be consistent with regressions (even test-only) introduced in 0.9.0, you can include it. On Wed, Jun 22, 2016 at 1:57 PM, Jean-Daniel Cryans <jdcry...@apache.org> wrote: > +1 > > I think we could grab > https://github.com/cloudera/kudu/commit/5f7d79910d8da492195b5c22a6d9a7911f950107 > and > https://github.com/cloudera/kudu/commit/679c6b2c93313bd4371916678b4ea8f52b621a50 > that fix issues committed in the master for 0.9.0. Adar can confirm if it's > desirable. > > This Java client fix wouldn't hurt: > https://github.com/cloudera/kudu/commit/d002e3257d49fe8e420c3d50eac54bb2d1952722 > > J-D > > > > On Wed, Jun 22, 2016 at 1:30 PM, Todd Lipcon <t...@cloudera.com> wrote: > >> Hey folks, >> >> Since 0.9 was released, we've run into a couple issues that have impacted >> both test clusters and a real-life cluster at JD.com: >> >> - a bug in tcmalloc-related metrics caused writes into already-freed memory >> which could cause lots of different crashes under load ( >> https://gerrit.cloudera.org/#/c/3445/) >> >> - a bug in Raft which could cause a tablet to get "stuck" and cease to make >> progress after a leader election (KUDU-1469, fix in review at >> https://gerrit.cloudera.org/#/c/3228/ ) >> >> The first of these in particular is a regression since 0.8 and is pretty >> critical (can cause crashes, incorrect data getting persisted, etc). So, I >> think we should try to address these issues with a patch release (0.9.1) >> ASAP. >> >> Thoughts? Anything else that we should consider as a candidate for 0.9.1? >> >> -Todd >> -- >> Todd Lipcon >> Software Engineer, Cloudera >>
