Re: [exim] heartbeat load balancing and exim

On Mon, Mar 31, 2008 at 03:32:16PM -0700, Jeroen van Aart wrote:
> > Exim works well with NFS provided that the client OS has decent support.
> > I've used Exim for delivery into NFS-based spools and it worked great
>
> I can imagine exim sort of working with NFS since it does not need
> things like file sharing and uses plain unix fileformat. I'd still avoid it.

It is properly supported by exim upstream, and the recommended course of
action in case you need a clustered setup. Exim deals correctly with the
multiple quirks that are part of the NFS protocol, such as certain
locking mechanisms not working (exim uses the one that does work on
NFS), the fact that the filesystem may not be available when exim tries
to write to disk (in which case exim will delay delivery, or return a
4xx response depending on what file it wants to write), and the fact
that multiple queue runners may be active at the same time (which is not
a problem even if you do it on the same host).

Exim does not 'sort of' work with NFS. Exim works great with NFS. Even
if you want to write to mboxes over NFS, exim will not corrupt them
provided you set your transport up to use lock files rather than flock
or fcntl locks. That doesn't mean using mbox on an NFS setup is a great
idea, but exim will do it, and do it properly, if you ask for it.

I can't see why the same would not be true for an IMAPd implementation.
NFS has its flaws; but if you implement your software properly, it's
perfectly possible to work around those. If you use an imapd that can do
Maildir or some such, I don't think there's any problem.

--
<Lo-lan-do> Home is where you have to wash the dishes.
-- #debian-devel, Freenode, 2004-09-22