----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/26373/#review57490 -----------------------------------------------------------
Thanks for the patch. Some comments below. core/src/main/scala/kafka/server/ReplicaManager.scala <https://reviews.apache.org/r/26373/#comment98208> This doesn't quite fix the original problem though. The original problem is that if the leader is not alive, we won't call partition.makeFollower(), in which the local replica is created. If a local replica is not created, the partition will be ignored when checkingpoint HW and we lose the last checkpointed HW. So, we have to call partition.makerFollower() for every follower, whether its leader is live or not. After this, we can proceed with the rest of the steps for only those partitions with a live leader. We can log a warning for those partitions w/o a live leader. - Jun Rao On Oct. 18, 2014, 7:26 a.m., Jiangjie Qin wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/26373/ > ----------------------------------------------------------- > > (Updated Oct. 18, 2014, 7:26 a.m.) > > > Review request for kafka. > > > Bugs: KAFKA-1647 > https://issues.apache.org/jira/browse/KAFKA-1647 > > > Repository: kafka > > > Description > ------- > > Addressed Joel's comments. > > > the version 2 code seems to be submitted by mistake... This should be the > code for review that addressed Joel's comments. > > > Diffs > ----- > > core/src/main/scala/kafka/server/ReplicaManager.scala > 78b7514cc109547c562e635824684fad581af653 > > Diff: https://reviews.apache.org/r/26373/diff/ > > > Testing > ------- > > > Thanks, > > Jiangjie Qin > >