Coda File System

Re: large servers: please help

From: Troy Benjegerdes <>
Date: Wed, 20 Jan 1999 13:45:33 -0600 (CST)
On 20 Jan 1999, Magnus Ahltorp wrote:

> > It's a good puzzle to see if Coda's connected semantics allow for the
> > atomic creation of a lock file. Perhaps that is just possible.  On the
> > other hand, I don't really have much more faith in AFS or NFS without lock
> > daemons when it comes to my mail.
> There are advisory locks in AFS, so mail handling would probably work,
> but AFS isn't designed for databases (I consider it a database if more
> than one mail is present in the file). I wouldn't recommend doing
> it. You're better off with MH mail boxes.

A month or two ago, I converted my personal procmail filtered Mail/
directory over to the qmail-style 'Maildir' format. I did this expecting
to move my Mail directory over to coda sometime, since I have several
machines at home. Pine (and the imapd provided with pine), procmail, mutt,
and a couple of other mail clients have patches for maildir style

maildir style mailboxes are safe for NFS, even though NFS does not provide
safe locking. Basically, each message is a separate file, and the file
name contains a timestamp and the message read/deleted/marked flags. There
are three subdirecories tmp/ new/ and cur/ in each mailbox folder. New
messages are written by the MDA in tmp/ and (atomicly?) moved to new/ when
writing is complete. The mail client then moves files from new/ to cur/
when read. 

Since maildir is nfs-safe (*without* using lockfiles), I believe it would
work quite well on Coda. However, since each message takes a file, you
will start eating up the 256000 file limit with a lot of users. If
conflicts occur, you might have 1 or 2 conflicting mail messages, but you
wouldn't lose the whole mailbox or ability to deliver mail.

| Troy Benjegerdes    |     |   |
|    Unix is user friendly... You just have to be friendly to it first.  |
| This message composed with 100% free software.   |
Received on 1999-01-20 14:44:32