[Gluster-devel] self heal problem

[Stephan von Krawczynski]

Hello,

please let me point you to a test case where we found out that the election of
the latest available file copy during replication self heal obviously does not
work out.

Take 2 servers and 2 clients with a standard replication setup.
The clients should have separate network paths to the servers, one link per
server.
Go to client 1 and ls -l some files
Now disconnect the network between client 2 and server 1 and rewrite these 
previsouly existing files to the glusterfs.
Make sure by ls -l on both servers that you see the files differing in size
and date.
Now go back to client 1 and ls -l the very same files.

We tried that with 2.0.9 and found out that merely half of the files get
self-healed wrong way round, the older file copies overwrite the newer ones.
How is that possible? Is there a way to check the relevant file infos to find
out why this happens?

-- 
Regards,
Stephan