> On Dec. 6, 2012, 10:41 p.m., Brock Noland wrote: > > flume-ng-sinks/flume-ng-hbase-sink/src/main/java/org/apache/flume/sink/hbase/HBaseSink.java, > > line 244 > > <https://reviews.apache.org/r/7657/diff/3/?file=177947#file177947line244> > > > > Why do we batch the puts into one runnable but not the increments? > > Hari Shreedharan wrote: > This is because Increments did not implement Row interface in 0.92.1 > release (which we use in the hadoop-1 profile), and so they cannot be batched
OK, perfect, I was not aware of this. Not being familiar with the HBase api, I still have one question, is there reason we don't send a list of Increments like we send a list of Rows in a single PutRunnable? - Brock ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/7657/#review14116 ----------------------------------------------------------- On Oct. 19, 2012, 12:14 a.m., Hari Shreedharan wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/7657/ > ----------------------------------------------------------- > > (Updated Oct. 19, 2012, 12:14 a.m.) > > > Review request for Flume. > > > Description > ------- > > Modify the Hbase sink to use multiple threads to wait on RPC calls. Since the > threads are blocked often, this will not cause too many running threads. > > > This addresses bug FLUME-1649. > https://issues.apache.org/jira/browse/FLUME-1649 > > > Diffs > ----- > > > flume-ng-sinks/flume-ng-hbase-sink/src/main/java/org/apache/flume/sink/hbase/HBaseSink.java > 021ecd0 > > Diff: https://reviews.apache.org/r/7657/diff/ > > > Testing > ------- > > All current unit tests pass. Did functional testing. > > > Thanks, > > Hari Shreedharan > >
