Quoting The Little Prince ([EMAIL PROTECTED]): > reading mboxes over NFS is always a bad idea. > really the only circumstance where you can use mail over nfs reliably is > if you use Maildir delivery instead of mbox.
For some definition of "Only" I <HEART> gross over generalizations. It would help if the original writer provides useful info. Sun and FreeBSD NFS clients prove really good solid. I've never had a NetApp problem. How many boxes read this set of mailboxes? How many users? What's going on? Now, to System 7 vs. Maildir (or, generically, file per message): Nick Christenson did a paper on Earthlink's mail system from 1997 or so. Sun's and NetApps. I did a variation of that with him and someone else at a mutual company later. Wasn't a problem. We did work around NFS locking by using our own locking system. But that was the big deal. A file that's been fsync'd and unlocked should appear as the same size to all mounts. Hacks to the getting mail and the mail.local which made this work. Nothing huge. Also, nothing shareable. (and I don't have them at this time.) One key way to avoid it is to not have shared (multiple machines) accessing the mailbox. But per NPC's "Sendmail Performance Tuning" book, he notes that System 7 mailboxes (mbox) are terribly inefficient. I'll note that putting 30k users on LOCAL disk will be FAR FAR faster than putting them on a network disk. Mail is all about I/O. I'll take a 500MHz Pentium for $600 and attach it to a really nice Baydel RAID for $15k and smoke a 4CPU box with cheapo drives. Spend the money right. A race car with model T tires won't help you win a race. Single file/message is also FAR more scalable and efficient. See also Cyrus, a solid IMAP server that's proven to scale very high. Courier which is close enough to IMAP for most mail clients (with annoying bumps) hasn't been around enough to really be brutalized. It looks like a promising youngster. > > We have been having a problem that was a bear to track down. > > > > When we use server mode we are losing mail. I know about the MDA and MUA > > issues and we are only using MDA for these accounts. If we turn on full > > debugging what we are seeing happen is that when a new message is delivered > > to the mail spool while a qpopper session is active qpopper at times does > > not realize that the mail spool has been appended and truncates the mail > > spool without first copying any undeleted messages.
