Warning |
---|
The wiki pages are not used for documentation any more. Please visit http://bookkeeper.apache.org for latest documentation. |
A comparison between BookKeeper and GlusterFS.
...
AFR replicates by write all changes to all child translators, much like RAID1. If a child is down, write will go to the other children. If all children go offline, the volume itself goes offline.
Consistency issues.
Imagine a system with a AFR translator replicating to two Posix translators (A & B). If one replica(A) goes offline, a client will continue to write to the other (B). Now, B goes offline. The client will stop writing to any replica. A is restarted so the client can continue writing. The client writes updates to A. Now A and B are inconsistent.
Differences to BookKeeper
1 http://www.gluster.com/community/documentation/index.php/GlusterFS_Concepts
2 http://www.gluster.com/community/documentation/index.php/Understanding_AFR_Translator