THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
Agenda:
- Status Updates
- Topics to discuss:
- Release plan for 4.8 ? new features ?
- Time for a 4.7.1 release ? 14 issues/PRs cherry-picked
- BP-14 status, how LastAddConfirmed advances ? (current PR https://github.com/apache/bookkeeper/pull/853 )
- option 1: advance LAC only force()
- option 2: advance LAC as regular non DEFERRED_SYNC mode
- option 3: option 1 + introduce write flag OPTIMISTIC_LAC for option 2
- BP-32 Advisory close (Charan)
- Status of multi entry loggers (Charan)
- Server side Backpressure (Andrey)
- BPs https://github.com/apache/bookkeeper/pulls?utf8=%E2%9C%93&q=is%3Aopen+bp+
- PRs https://github.com/apache/bookkeeper/pulls
Attendees:
Sam, Sijie, Enrico, JV
Minutes:
- Sijie, we have still a lot of time. Let's see the list of feature in July when we are going to cut the release branch.
- JV: we do not yet have a comprehensive coverage on testing to make us conforrtable with releases
- BP-14, let's sttick with option1, Sijie motivation for option 2 is useful in Clod environment, but JV explains that without the "sync" barrier of durable LAC all the stack will be confused (think about the replicator),
- JV: liveness of the Bookie: there is no way to detect bookie failure only using ZK session
- Enrico: we can do external checks
- Enrico talked about the change about SortedLedgerStorage vs InterleavedStorage, It is fine to change the class Hierarchy
- JV talks about the proposal on PlacementPolicy (https://github.com/apache/bookkeeper/pull/1397). JV does not want a new 'tunable', but as Cloud environment there is no much difference in "racks", so OK for having an explicit opt in flag for this feature
- Enrico reported his recent problem with a Bookie which is alive but does not advertise itsselt on ZooKeeper, maybe it is due to use of ACLs