We have had numerous problems dealing with SunOS and Solaris mailtool file locking after moving our mail spool off our Solaris server. Initially, we moved the spool to a Network Appliance and used a Digital Unix system as the mail server. In an attempt to solve the locking problem, we moved the spool to a local filesystem on the Digital Unix box, but we still had problems. When we switched from binmail to procmail as our mail delivery agent, most of the problems were resolved.

In all of the mess, we learned a couple of things to look for: Is there a cron job on the Sun that clears locks older than a few minutes? If so, you might need a similar workaround on the NetApp filesystem. Are your users exiting OpenWindows without clicking "Done" (to save changes)? It is a bug to leave the file locked, but I believe that some versions of mailtool do this. Do your users find that their mailtools get confused? We found that to be a symptom of locking problems. Do your users have large inboxes? Our mailtool and dtmail users have a tendency to keep lots of old mail in their inboxes, and the resultantly large mail files take longer to update. This increases the likelihood of two processes needing to lock the file at the same time.

You might also want to rebuild procmail. When you build it, one thing it does is to test locking in your mail spool directory, so that it can determine what locking mechanisms work reliably. It may be that it needs to run those tests again on your new filesystem.

On the other hand, you might want to switch to pop3 or imap4 (if possible) to get mail to your users' workstations. These seem to be more robust protocols than nfs with locking.

Good luck. Dealing with mail problems can be unpleasant.

David Ritch