On Fri, Oct 17, 2003 at 03:59:22PM +0100, Mike Williams wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Friday 17 October 2003 15:49, Matthias F. Brandstetter wrote:
Hmm.. Ok, that's a good point. But then I still have the problem with the central mail storage :(
What's wrong with NFS?
Hell, I've run a system with 16 machines doing mail all sharing an NFS export and mbox's!
It doesn't scale (with mbox) to hundreds of machines, and many implementations of NFS on non-linux systems don't implement file-locking properly, which especially hurts mbox.
On the systems I manage at work we share mail (with-mbox) to over 130 machines via NFS without breaking a sweat. We support another several hundred users via imap and pop. Broken NFS is largly a thing of the past (or non Sun/Linux/BSD Unix).
Doing maildir over NFS is a perfectly valid way to do this, but you end up with at least three boxes... a highly redundant NFS file server and mail servers tied to it.
Or just run Sun hardware (with solaris)... If a CPU/memory module fails it is simply turned off and the system continues to work. And like any modern OS, use Raid for the mail. Run the raid off of an external storage rack/case. If you machine explodes unplug raid box and plug into backup. Keeping mailserver config files synced between 2 boxes is simple.
I don't understand why mail can't be offline for several hours while a server is repaired? Automatic failover is more trouble than it's worth when all were talking about is mail.
One big box isn't such a bad thing... Just keep spare parts around in case something goes wrong (or an entire spare machine). Always use raid for important data (like mail) along with regular backups.
-- Bryan Whitehead Email:[EMAIL PROTECTED] WorkE:[EMAIL PROTECTED]
-- [EMAIL PROTECTED] mailing list
