[Dovecot] Dovecot v1.1.alpha4 problem
Dovecot v1.1.alpha4 imap after some time stops responding. It can not be restarted as /etc/init.d/dovecot restart
killall imap also does not help
killall -9 imap and then restart does help
in log I see message (repeted periodically) that Sep 12 09:50:02 localhost postfix/local[22533]: C7057753B6: to=<username@domain> , relay=local, delay=5195, delays=5156/0.01/0/39, dsn=4.2.0, status=deferr ed (cannot update mailbox /home/username/Mailbox for user username. unable to lo ck for exclusive access: Resource temporarily unavailable)
lsof -n shows that the file is is open by imap process.
The problem persists with mbox inbox folders in v1.1.alpha4. I get a stalled process "imap" which can not be killed by killall imap only by killall -9 imap This process takes a lock of user mailbox thus prevents the user to read/send mail. Other users are OK
Vladislav Malyshkin wrote:
Dovecot v1.1.alpha4 imap after some time stops responding. It can not be restarted as /etc/init.d/dovecot restart
killall imap also does not help
killall -9 imap and then restart does help
in log I see message (repeted periodically) that Sep 12 09:50:02 localhost postfix/local[22533]: C7057753B6: to=<username@domain> , relay=local, delay=5195, delays=5156/0.01/0/39, dsn=4.2.0, status=deferr ed (cannot update mailbox /home/username/Mailbox for user username. unable to lo ck for exclusive access: Resource temporarily unavailable)
lsof -n shows that the file is is open by imap process.
On Thu, 2007-09-13 at 00:48 -0400, Vladislav Malyshkin wrote:
The problem persists with mbox inbox folders in v1.1.alpha4. I get a stalled process "imap" which can not be killed by killall imap only by killall -9 imap This process takes a lock of user mailbox thus prevents the user to read/send mail.
Could you get gdb backtrace of those processes (a few times if they're different)?
gdb attach <imap process pid> bt
(if attach doesn't work, replace it with path to imap binary)
Timo, this is very hard for me to do. Every time this problem starts people get panicked, immediately call me, call others. This did happen twice in the last 24 hours. Typically most active user gets his mbox locked after 10-12 hours. I tried to reproduce the problem on another mail server (where a problem does not make people panic and call me at night) - no luck, because this server does not have such a large mail traffic. I downgraded main server to dovecot 1.0.5. Looks OK for now. I can send you /var/log/maillog but there is no much information there, only about postfix not able to get exclusive lock of user mailbox. Also, this may somekind help, the imap process taking lock can not be killed by -15 signal, only by -9. Vladislav
Timo Sirainen wrote:
On Thu, 2007-09-13 at 00:48 -0400, Vladislav Malyshkin wrote:
The problem persists with mbox inbox folders in v1.1.alpha4. I get a stalled process "imap" which can not be killed by killall imap only by killall -9 imap This process takes a lock of user mailbox thus prevents the user to read/send mail.
Could you get gdb backtrace of those processes (a few times if they're different)?
gdb attach <imap process pid> bt
(if attach doesn't work, replace it with path to imap binary)
On Thu, 2007-09-13 at 07:37 -0400, Vladislav Malyshkin wrote:
this is very hard for me to do. Every time this problem starts people get panicked, immediately call me, call others. This did happen twice in the last 24 hours.
Maybe you should not test alpha software in production? :D
ciao
Luca
participants (3)
-
Luca Corti
-
Timo Sirainen
-
Vladislav Malyshkin