Autor: Robin Kearney Data: A: exim-users Assumpte: Re: [exim] Shared Spool using NFS
On 5 Sep 2005, at 12:20, Tony Finch wrote:
> On Sun, 4 Sep 2005, Xirtam wrote:
>
>>
>> My question: Are there technical problems when more than one Exim
>> instance is
>> using the same spool directory? Are there workarounds or *if*
>> there are
>> problems could they be solved without losing the idea of a
>> redundant mail
>> system with a central storage unit?
>>
>
> They have used a setup like this at demon.nl. You need to look at the
> localhost_number setting which prevents multiple Exim's from
> generating
> clashing message IDs. It's an expanded string, so you can set it
> based on
> your hostname (say) in order to keep otherwise identical
> configurations on
> each Exim. You should also ensure that $spool_directory/db is
> unique per
> host, because of locking problems (it would cause a cluster-wide
> bottleneck, and in any case distributed locking protocols are good to
> avoid).
I've been testing a HA failover rig with the exim spool on a drbd
volume. This
setup seems to work very well although only one exim is active at any
one time.
You take a hit in disk performance because of the drbd network
transfer but with
a dedicated Gb network this seems not to impact the performance to much.