Matt Prigge <[EMAIL PROTECTED]> wrote:
> > Right, the 'mailstore-entities' (actually a cluster of two
> > machines sharing a raid-array) have to be redundant
> > themselves - as it was mentioned by Atif Ghaffar already.
>
> Is that even possible? I dont doubt your suggestion, I just didnt know that
> linux would support sharing the same array with n+1 machines. I was pretty
> sure that there would be file locking issues or something along those lines.
> Are there any specific restrictions on this? (type of FS used, type of
> array, etc) Im guessing that there is probably a website somewhere devoted
> to this topic, but if there is I havent found it yet.
Maybe Linux-HA is a good start (http://linux-ha.org/).
> I have toyed with the idea of using a shared
> fiberchannel SAN for a long time, but I was pretty sure
> it wouldnt work (not to mention they are sort of
> expensive.. hehe.).
Which is basically the reason I haven't tested that as
well. There is a product called 'Reliant Monitor Software
(RMS)' from Fujitsu-Siemens, which they ported to Linux
for their Primergy PC-Servers. It includes some software
called DuplexDataManager - DDM
http://manuals.mchp.siemens.de/infothek/dprimswea/BOOKS/ \
DuplexDataManager_e1/duplexdatamanager_e1.pdf),
which is able to write to two seperated disk-arrays via
FCAL (I think).
What I meant by 'sharing' a raid-array was really 'are
attached to storage, which holds the mailboxes'.
They don't need to write to the same disks at the same
time. You just have to be able to mount the disksets from
the standby-machine, after the live-machine has crashed.
by
Töns
--
Linux. The dot in /.
