atom feed6 messages in net.sourceforge.lists.courier-maildropRe: [maildropl] Maildrop and Filesyst...
FromSent OnAttachments
Nathan FrankishNov 29, 2007 2:01 am 
Tony EarnshawNov 29, 2007 3:43 am 
Nathan FrankishNov 29, 2007 3:56 am 
Jay LeeNov 29, 2007 5:18 am 
Nathan FrankishNov 29, 2007 1:30 pm 
Jay LeeNov 29, 2007 8:04 pm 
Subject:Re: [maildropl] Maildrop and Filesystem quotas..
From:Nathan Frankish (
Date:Nov 29, 2007 1:30:36 pm

You are correct. The file system quota is in place to look after the entire domain (including www/ftp). in terms of how i decided to setup, i basicly allowed it that they could overalocate quota for mailboxs (ie if the domain has 1gig enforced by the file system, they could still have 100 mailboxes with a mail quota of 100mb) but that they would get stopped by the file system. It does seem that it isnt posslbe to make it play nice, so i shall work with it as is.

Thanks for the asstaince.


Jay Lee wrote:

On 11/29/07, *Nathan Frankish* < <>> wrote:

So within my setup there is both file system quotas, and individual mail quotas (from mysql). The mail quotas, maildrop handles fine and returns a reasonal message when the account has hit the end. However when the file system quota runs out, all it gets back is unable to deliver to mailbox..

Why both? Is the FS quota per user or for a group of users / domain? FS quota are hard, they are in no way flexible. The error the kernel returns when a program tries to write past a quota is pretty much the same thing as trying to write to a disk that is full. AFAIK, it's not easy to determine how much space a FS quota has left or gracefully recover when the quota is hit. All of these reasons are why Courier has soft quotas to begin with. But of course, soft quotas aren't as strict as some admins prefer, each has their positives and negatives. What if you set the FS quota say 5% higher than the soft quota? Then users would get the graceful warnings of the soft quota (see the maildrop -w option) and if they squeeze their way past the soft quota, they hit the hard FS quota...


My understand was that the filesystem would have obviously told maildrop that it cant write antyhing, and more then likely said that its out of quota.. but maildrop either ignores that information, or isnt setup to recieve it properly.. Im just wondering what else i can do, short of making a maildropfilter that checks it first, to make it play nice with the file system and return a decent error if it is unable to deliver due to file system quota exhausted..

Thanks! Tony Earnshaw wrote: > Nathan Frankish skrev, on 29-11-2007 11:01: > > >> During my testing of maildrop with file system quotas, i noticed that >> when the virtualuser was out of space (due to hitting a file system >> quota) that the error message returned was quite ambiguous >> >> Nov 29 09:44:23 leela postfix/pipe[3585]: 3FFF78ADD6: >> to=< <> >, relay=maildrop, delay=26, delays=23/1/0/1.5, >> dsn=4.3.0, status=deferred (temporary failure. Command output: maildrop: >> error writing to mailbox. /usr/bin/maildrop: Unable to deliver to mailbox. ) >> >> Im wondering if that error is displayed instead of a simple quota >> reached issue or something due to not being able to make a dotlock file >> or something simular, but i really have no clue.. Any suggestions on how >> to make it return a decent error that wouldnt be confusing if the email >> eventually bounced? >> > > What does maildirsize (file) in the user's Maildir say about the quota > when that happens? It sounds like maildrop isn't getting the quota anywhere. > > I use maildrop with LDAP and have my quotas there - maildrop gives a > bounce back to Postfix (permission denied. Command output: maildrop: > maildir over quota.). Postfix (bounce) then sends a bounce notice back > to the sender. Besides which, that method is far more flexible: quotas > can more easily be changed than by using file system quotas - we have > frequently to increase quotas on our rigs. > > Best, > > --Tonni > >