> On Sept. 13, 2012, 7 p.m., Brock Noland wrote: > > flume-ng-channels/flume-file-channel/src/main/java/org/apache/flume/channel/file/EventQueueBackingStoreFile.java, > > line 105 > > <https://reviews.apache.org/r/7086/diff/1/?file=154337#file154337line105> > > > > I don't see the sync() call? Why not keep the call to elementsBuffer > > and then use it's force() method? > > Hari Shreedharan wrote: > Note that checkpointFileHandle is a RandomAccessFile. RandomAccessFile > does not need a sync call, does it? The write call will only write the 8 > bytes - while I could not find documentation which explains what a > byteBuffer.force() forces.
With a RAF you either have to call getFD().sync() http://docs.oracle.com/javase/6/docs/api/java/io/FileDescriptor.html#sync() or getChannel().force() http://docs.oracle.com/javase/6/docs/api/java/nio/channels/FileChannel.html#force(boolean) to tell the OS to sync the pages to disk. It as the same affect as http://docs.oracle.com/javase/1.4.2/docs/api/java/nio/MappedByteBuffer.html#force() - Brock ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/7086/#review11490 ----------------------------------------------------------- On Sept. 13, 2012, 6:54 p.m., Hari Shreedharan wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/7086/ > ----------------------------------------------------------- > > (Updated Sept. 13, 2012, 6:54 p.m.) > > > Review request for Flume. > > > Description > ------- > > CHECKPOINT_INCOMPLETE will now be written synchronously to disk before > starting checkpoint. > > > This addresses bug FLUME-1576. > https://issues.apache.org/jira/browse/FLUME-1576 > > > Diffs > ----- > > > flume-ng-channels/flume-file-channel/src/main/java/org/apache/flume/channel/file/EventQueueBackingStoreFile.java > 4717055 > > Diff: https://reviews.apache.org/r/7086/diff/ > > > Testing > ------- > > > Thanks, > > Hari Shreedharan > >
