----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/19088/#review36898 -----------------------------------------------------------
Ship it! - Jay Kreps On March 12, 2014, 1:50 a.m., Jun Rao wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/19088/ > ----------------------------------------------------------- > > (Updated March 12, 2014, 1:50 a.m.) > > > Review request for kafka. > > > Bugs: KAFKA-1301 > https://issues.apache.org/jira/browse/KAFKA-1301 > > > Repository: kafka > > > Description > ------- > > 1. Fixed the unsent check in Sender.run(). 2. Sometimes, the test fails since > some messages failed all retries. The issue is that the config for > replica.fetch.min.bytes is too large. This causes many produce requests > (ack=-1) to have to wait for replica.fetch.wait.max.ms=500ms to complete. If > a metadata request is queued up behind those produce requests, the producer > may not see the current leader and therefore fails all tries. > > > Diffs > ----- > > > clients/src/main/java/org/apache/kafka/clients/producer/internals/RecordAccumulator.java > 616e1006f21c54af8260e84a96928cb8893ceb7c > > clients/src/main/java/org/apache/kafka/clients/producer/internals/Sender.java > 2acb96df9f4d5c7714a752517d431d16c48dcf1d > system_test/replication_testsuite/config/server.properties > c6284122e3dfaa046a453511c17a19a536dc1035 > > Diff: https://reviews.apache.org/r/19088/diff/ > > > Testing > ------- > > > Thanks, > > Jun Rao > >