Re: [Exim] NFS mounted/shared queue

<-- Warning: charset 'format' is not supported -->

Quoting Suresh Ramasubramanian <linux@???>:

> How does the solution that was implemented deal with nfs locking issues?

The version of Exim was patched to deal with NFS locking issues. I have the
original source, so doing a diff I can see various locking related code
snippets inserted to deal with this issue. Exim is built on each server so
that the 'db' and 'lock' directories are local to the box, whilst the 'msglog'
and 'input' directories are symlinks to an NFS mount (/queue). It is the
msglog and input directories which are shared across all 3 mail servers.

>> PS. I am in the mount locally and don't share a mail queue camp. I'm just
>> wondering what drove my predecessors to do it this way!
>
> So am I. Mounting maildirs over NFS (using a SAN / NAS) is a good idea. Not
> so mounting spools.

Interesting enough, we haven't had any issues with the current solution. I
believe NFS+Linux has come a long way in the past couple of years and is very
stable these days. I don't have any issues with mounting over NFS - I am more
worried about sharing queues with other mail servers.

I am *very* reluctant to continue with a patched Exim, so are we saying
that an
unpatched Exim should not have it's queue mounted over NFS and that this will
cause locking issues in the future? Even if we have a dedicated NFS queue
mount per server?

Cheers,
Ray