On Wed, Aug 6, 2008 at 2:56 PM, Alexander Bokovoy <ab@xxxxxxxxx> wrote:
> On Wed, Aug 6, 2008 at 7:28 AM, ronnie sahlberg
> <ronniesahlberg@xxxxxxxxx> wrote:
>> CTDB/Samba has never been tested with and havent been designed for use
>> with with
>> "lockless" cluster filesystems, so there are probably a lot of
>> things that can (and probably will) break in subtle and
>> hard-to-diagnose ways.
>> Good luck, and keep us informed.
> What about cases where CTDB is performing HA function without actual
> CIFS/NFS use? Being a heart-beat to web or other services?
Sure you can use ctdb as a active/passive failover hearbeat solution
But since ctdb is designed to be an all-active solution for CIFS it
may be a bit overkill to
use a quite complex solution like ctdb to just use it as a heartbeat
for active/passive failover.
(it is very easy to go from all-active to active/passive since the
latter can be considered just a special subset of the former while it
may be very difficult to go in the other direction)
> Will it
> work with a CIFS/NFS mount to share a lock for preventing brain split?
Not having a cluster filesystem backend I would work either by using a
CIFS or an NFS mount
as you mention.
I would personally use a CIFS mount since byte range locking is A LOT
more reliable and robust
than in NFS locking.
NFS locks working-correctly,
pick two. you cant have all three. :-)
> / Alexander Bokovoy