I am also running vpopmail-5.2.1. I have replaced vdelivermail in all .qmail-default for all domains with maildrop (1.5.2) and it seems to work a little better. So far, after removing all old maildirsize files and recreating them, I have not seen the problem again.
One way things can get screwed up is if you have aliases that point to a user's maildir directly (as added by qmailadmin). This way, any mail sent to the alias, will be delivered to the Maildir of the user by qmail-local. I have suspicions that the maildir++ patch is buggy, either doesn't recalculate the maildirsize file at all, or just doesn't add the messages. I am not sure if this is related to NFS, it very well could be. Do you offer any sort of quota statistics to the end users? For example, if you offer a graphical quota bar, or just quota usage text to your end users, you'll probablly get some calls which will give you a better idea of how often this breaks. Another good way to check, instead of having your users find out for theirselves, is to write a script that gets a list of all mailboxes (by reiterating through each domain directory in domains) then runs `~vpopmail/bin/vuserinfo -Q [EMAIL PROTECTED] for every email account and checks if this returns 0% usage while the user's Maildir size is more then 1M as reported by `du -sh ~vpopmail/domains/domain.com/username/Maildir`. I already have a shell script that does that, and this is my only way of find out any occurunces of the negative quota usage problem, without logging in to every email account via imap and issuing a 'getquota ROOT' command to see if I get negative value back for storage usage. If you need this script, let me know.. Respectfully, Tim Hasson Quoting Nick Harring <[EMAIL PROTECTED]>: > I've got about 50K boxes on NFS with maildir++. I'm running vpopmail > 5.2.1 though. The only problem I've ever seen is every once in a great > while Courier-IMAP will return a wildly wrong negative value for number > of unread messages, which breaks some other stuff I run, but I doubt > there's a correlation. > > Cheers, > Nick Harring