jon@??? said:
} Unfortunately this assumption doesn't scale particularly well :(
Actually I think it scales *much* better than the single filesystem model.
Where it has problems is that if you lose a spool disk, you have lost the
mail on it - I am afraid we don't run on RAID disks for the exim spool.
Worse, its close to impossible to back it up in any sensible way.
Personally, from bitter experience, I don't trust NFS locking at all. I
also try to restrict NFS ops to the ones the system does well - directory
lookups, block reads, block writes - this latter really causes problems
with DB files which I find build on NFS filesystems more 20 times slower
than on local disks (so where I put DB (read only) files on NFS I build on
local and then copy).
This means I run with a set of machines (currently 3), sharing an NFS user
spool (which is designed in a way that negates the use of file locking),
and with local exim spool, config and log directories (I process the logs
into a central location overnight).
BTW all our mail machines have the same config, all run with a similar
load average and number of exim processes. All are referred to
identically in MX records - they appear as a single name with multiple IPs
in them all. However one always runs with a 50% smaller queue than the
others.... and I don't know why!
Nigel.
--
[ Nigel.Metheringham@??? - Systems Software Engineer ]
[ Tel : +44 113 251 6012 Fax : +44 113 224 0003 ]
[ Friends don't let friends use sendmail! ]
--
* This is sent by the exim-users mailing list. To unsubscribe send a
mail with subject "unsubscribe" to exim-users-request@???
* Exim information can be found at http://www.exim.org/
