[Gluster-devel] Effect of AFR's known file re-open issue on MySQL

[David Sickmiller]

Hi,

I'd like to confirm that I'm properly understanding how the "File 
re-opening after a server comes back up" known AFR issue affects the use 
of MySQL/InnoDB (and probably any other RDBMS).  My impression is that 
MySQL/InnoDB opens a bunch of files when it starts, and it generally 
holds these files open until the database is shut down.  This would mean 
that, if MySQL was started while a mirrored brick is offline, 
replication wouldn't happen until after shutting down MySQL.  Is that 
so?  This would unfortunately seem to work against my reason for having 
a redundant server cluster.

Is there any way to mitigate this issue?  For example, would running "ls 
-lR" on the rejoining brick sync it up?

If not, any guesses on when the fix will be available?

Thanks,
David