[Dovecot] distributed mdbox

[Stan Hoeppner]

On 3/22/2012 11:17 AM, Jim Lawson wrote:
> On 03/22/2012 12:11 AM, Stan Hoeppner wrote:
>> On 3/21/2012 12:04 PM, Timo Sirainen wrote:
>>> The problem is most likely the same as with NFS: Server A caches data -> server B modifies data -> server A modifies data using stale cached state -> corruption. Glusterfs works with FUSE, and FUSE has quite similar problems as NFS.
>>>
>>> With director you guarantee that the same mailbox isn't accessed simultaneously by multiple servers, so this problem goes away.
>> If using "real" shared storage i.e. an FC or iSCSI SAN LUN, you could
>> use a true cluster file system such as OCFS or GFS.  Both will eliminate
>> this problem, and without requiring Dovecot director.  And you'll get
>> better performance than with Gluster, which, BTW, isn't really suitable
>> as a transactional filesystem, was not designed for such a use case.
> 
> Speaking as an admin who has run Dovecot on top of GFS both with and
> without the director, I would never go back to a cluster without the
> director.  The cluster performs *so* much better when glocks can be
> cached on a single node, and this can't happen if a single user has IMAP
> processes on separate nodes.
> 
> No, you don't strictly need the director if you have GFS, but if you can
> manage it, you'll be a lot happier.

Did/do you see the Director/glock benefit with both maildir and mdbox
Jim?  Do you see any noteworthy performance differences between the two
formats on GFS, with and without Director?  BTW, are you hitting FC or
iSCSI LUNs?

-- 
Stan