Re: [Dovecot] Catch22: user needs space to fix out of space condition
Thanks to all who've made suggestions. It seems removing dotlocks as a locking method is the way to go. There is another dotlock locking variant mentioned in 10-mail.conf that seems to address this situation for those that can't get away from dotlocks:
# dotlock_try: Same as dotlock, but if it fails because of permissions or
# because there isn't enough disk space, just skip it.
mbox_write_locks = dotlock_try fcntlJoseph Tam jtam.home@gmail.com
On Fri, 26 Aug 2011, Joseph Tam wrote:
Thanks to all who've made suggestions. It seems removing dotlocks as a locking method is the way to go.
Actually, this gives me pause that maybe I should not enirely remove the dotlocking method
http://mailman2.u.washington.edu/pipermail/alpine-info/2008-July/000996.htmlAny comments on the (sole) use of POSIX fcntl() type locking?
Joseph Tam jtam.home@gmail.com
On Fri, 2011-08-26 at 18:30 -0700, Joseph Tam wrote:
On Fri, 26 Aug 2011, Joseph Tam wrote:
Thanks to all who've made suggestions. It seems removing dotlocks as a locking method is the way to go.
Actually, this gives me pause that maybe I should not enirely remove the dotlocking method
http://mailman2.u.washington.edu/pipermail/alpine-info/2008-July/000996.html
Any comments on the (sole) use of POSIX fcntl() type locking?
As long as you haven't used symlinks in your mboxes there's no problems with fcntl locking with Dovecot (assuming there are no non-Dovecot software writing to them).
On Mon, 29 Aug 2011, Timo Sirainen wrote:
Actually, this gives me pause that maybe I should not enirely remove the dotlocking method
http://mailman2.u.washington.edu/pipermail/alpine-info/2008-July/000996.html
Any comments on the (sole) use of POSIX fcntl() type locking?
As long as you haven't used symlinks in your mboxes there's no problems with fcntl locking with Dovecot (assuming there are no non-Dovecot software writing to them).
Thanks for the info.
I don't meet the last criteria: some users have direct file access via pine and other mail readers. There's also procmail, but I don't know what locking method it uses. Reading the pine sources closely, I think it fakes fcntl() for NFS mailboxes.
I'm using dotlock_try which solves my immediate problem of allowing users to delete Email under full quota.
The long term solution is to replace file access with kerberized IMAP access, but that's much further down the road.
Joseph Tam jtam.home@gmail.com
Monday, August 29, 2011, 4:32:55 AM, Joseph wrote:
On Mon, 29 Aug 2011, Timo Sirainen wrote:
Actually, this gives me pause that maybe I should not enirely remove the dotlocking method
http://mailman2.u.washington.edu/pipermail/alpine-info/2008-July/000996.htmlAny comments on the (sole) use of POSIX fcntl() type locking?
As long as you haven't used symlinks in your mboxes there's no problems with fcntl locking with Dovecot (assuming there are no non-Dovecot software writing to them).
Thanks for the info.
I don't meet the last criteria: some users have direct file access via pine and other mail readers. There's also procmail, but I don't know what locking method it uses. Reading the pine sources closely, I think it fakes fcntl() for NFS mailboxes.
Pine/Alpine knows IMAP. That's what I use locally on my server.
I'm using dotlock_try which solves my immediate problem of allowing users to delete Email under full quota.
The long term solution is to replace file access with kerberized IMAP access, but that's much further down the road.
Joseph Tam jtam.home@gmail.com
-- Best regards, Duane mailto:duane@duanemail.org
participants (3)
-
Duane Hill
-
Joseph Tam
-
Timo Sirainen