[Dovecot] dovecot stopping - no messages
Greetings
Running Fedora Core 3 and dovecot-0.99.13-3.FC3 as a production mail server for around 30+ accounts.
A couple of times a day dovecot just stops.
Nothing in the mail log except for
Apr 26 00:48:19 sydsrv05 imap-login: Can't connect to auth process at default: Connection refused
It's got to the point that I have put a script into cron.hourly to restart the dovecot service every hour.
Any suggestions where to start looking please?
Rgds
Nigel
-- Nigel Allen, Snr Consultant, ProSmart Consulting Australia 48 Tiarri Avenue, Terrey Hills, NSW 2084, Australia Tel: +61 2 9450 2690 Fax: +61 2 9450 2691 Mob: +61 4 1494 5269 http://www.prosmart.com.au
On Tue, 2005-04-26 at 12:32 +1000, Nigel Allen wrote:
Greetings
Running Fedora Core 3 and dovecot-0.99.13-3.FC3 as a production mail server for around 30+ accounts.
A couple of times a day dovecot just stops.
Nothing in the mail log except for
Apr 26 00:48:19 sydsrv05 imap-login: Can't connect to auth process at default: Connection refused
Sounds like Dovecot's master process crashes. Is there a core file in /var/run/dovecot directory? If so, do:
gdb dovecot /var/run/dovecot/core bt
and send the result here.
Or alternatively do:
gdb attach pidof dovecot
cont
(wait for it to crash)
bt
Hi,
Thanks for the prompt reply.
Astonishingly, I came into the clients this morning and the damned thing had not fallen over once.
I suspect the problem may have been imap related. I'll explain.
The machine was an upgrade from what was essentially a pop3 RH9 installation. We installed both the imap and pop features of dovecot in order to continue the email to users (all set up to do pop3 at the moment) and to test the imap features. The reason we left the pop stuff alone is that the mailer they use on their windoze boxen grabs the mail and stores it in it's own mailbox format.
The only thing I did yesterday was to take the lines referencing imap out of the dovecot.conf and restart the service. It's been running like a train ever since.
I can't spare time to investigate this now but will try it again once the "busy" period of the switchover has happened.
Thanks again for the prompt response - hell of a product btw.
Rgds
Nigel.
On Tue, 2005-04-26 at 19:13, Timo Sirainen wrote:
On Tue, 2005-04-26 at 12:32 +1000, Nigel Allen wrote:
Greetings
Running Fedora Core 3 and dovecot-0.99.13-3.FC3 as a production mail server for around 30+ accounts.
A couple of times a day dovecot just stops.
Nothing in the mail log except for
Apr 26 00:48:19 sydsrv05 imap-login: Can't connect to auth process at default: Connection refused
Sounds like Dovecot's master process crashes. Is there a core file in /var/run/dovecot directory? If so, do:
gdb dovecot /var/run/dovecot/core bt
and send the result here.
Or alternatively do:
gdb attach
pidof dovecot
cont (wait for it to crash) btNigel Allen, Snr Consultant, ProSmart Consulting Australia 48 Tiarri Avenue, Terrey Hills, NSW 2084, Australia Tel: +61 2 9450 2690 Fax: +61 2 9450 2691 Mob: +61 4 1494 5269 http://www.prosmart.com.au
Re my last "everything is fine" message.
Yup, you guessed it - about 5 minutes ago the users all started ringing me - mail had died again.
Tried a "service dovecot status" and got a "dovecot dead but subsys locked". Did a "service dovecot start" and away it went again.
From the earlier advice, there is no core in /var/run/dovecot but I will try the gdb instructions (newbie with gdb) and post the resuults.
Thanks again
Nigel.
On Tue, 2005-04-26 at 19:13, Timo Sirainen wrote:
On Tue, 2005-04-26 at 12:32 +1000, Nigel Allen wrote:
Greetings
Running Fedora Core 3 and dovecot-0.99.13-3.FC3 as a production mail server for around 30+ accounts.
A couple of times a day dovecot just stops.
Nothing in the mail log except for
Apr 26 00:48:19 sydsrv05 imap-login: Can't connect to auth process at default: Connection refused
Sounds like Dovecot's master process crashes. Is there a core file in /var/run/dovecot directory? If so, do:
gdb dovecot /var/run/dovecot/core bt
and send the result here.
Or alternatively do:
gdb attach
pidof dovecot
cont (wait for it to crash) btNigel Allen, Snr Consultant, ProSmart Consulting Australia 48 Tiarri Avenue, Terrey Hills, NSW 2084, Australia Tel: +61 2 9450 2690 Fax: +61 2 9450 2691 Mob: +61 4 1494 5269 http://www.prosmart.com.au
Okay - after 3 hours I got this:
Detaching after fork from child process 31368. Detaching after fork from child process 31371. Detaching after fork from child process 31372. Detaching after fork from child process 31376. Detaching after fork from child process 31377. Detaching after fork from child process 31379. Detaching after fork from child process 31380. Detaching after fork from child process 31382. Detaching after fork from child process 31383. Detaching after fork from child process 31385. Detaching after fork from child process 31386. Detaching after fork from child process 31388. Detaching after fork from child process 31389. Detaching after fork from child process 31391. Detaching after fork from child process 31392. Detaching after fork from child process 31394. Detaching after fork from child process 31395. Detaching after fork from child process 31397. Detaching after fork from child process 31398. Detaching after fork from child process 31400. Detaching after fork from child process 31401. Detaching after fork from child process 31403.
Program received signal SIGSEGV, Segmentation fault. 0x0804fb3a in ?? () (gdb) bt #0 0x0804fb3a in ?? () #1 0x09bd1960 in ?? () #2 0x09bd197c in ?? () #3 0xbfee1e98 in ?? () #4 0x0804a203 in ?? () #5 0x09bd1ad8 in ?? () #6 0x00000a42 in ?? () #7 0x09bd26d8 in ?? () #8 0x08051049 in ?? () #9 0xbfee1ed0 in ?? () #10 0x00000000 in ?? () (gdb) q The program is running. Quit anyway (and detach it)? (y or n) Detaching from program: /usr/sbin/dovecot, process 18560
Any ideas please?
TIA
Nigel.
On Wed, 2005-04-27 at 08:55, Nigel Allen wrote:
Re my last "everything is fine" message.
Yup, you guessed it - about 5 minutes ago the users all started ringing me - mail had died again.
Tried a "service dovecot status" and got a "dovecot dead but subsys locked". Did a "service dovecot start" and away it went again.
From the earlier advice, there is no core in /var/run/dovecot but I will try the gdb instructions (newbie with gdb) and post the resuults.
Thanks again
Nigel.
On Tue, 2005-04-26 at 19:13, Timo Sirainen wrote:
On Tue, 2005-04-26 at 12:32 +1000, Nigel Allen wrote:
Greetings
Running Fedora Core 3 and dovecot-0.99.13-3.FC3 as a production mail server for around 30+ accounts.
A couple of times a day dovecot just stops.
Nothing in the mail log except for
Apr 26 00:48:19 sydsrv05 imap-login: Can't connect to auth process at default: Connection refused
Sounds like Dovecot's master process crashes. Is there a core file in /var/run/dovecot directory? If so, do:
gdb dovecot /var/run/dovecot/core bt
and send the result here.
Or alternatively do:
gdb attach
pidof dovecot
cont (wait for it to crash) btNigel Allen, Snr Consultant, ProSmart Consulting Australia 48 Tiarri Avenue, Terrey Hills, NSW 2084, Australia Tel: +61 2 9450 2690 Fax: +61 2 9450 2691 Mob: +61 4 1494 5269 http://www.prosmart.com.au
participants (2)
-
Nigel Allen
-
Timo Sirainen