[Dovecot] Dovecot dies, maybe ntpdate related. I'm new to dovecot
Hi all, my first post.
After years with an old server, a couple of months ago I've installed a new one.
We use IMAP. After research, I chosed dovecot.
It has been running for several weeks but from one week ago or so, it suddently dies about once per day.
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each night. We did this because "apparently" we did not a "soft time drift" installing the full ntp daemon, but I start thinking the time-jump crashes dovecot.
QUESTIONS:
- Might this time-jump cause dovecot to die?
- I don't know wehere the dovecot logs are written to. Where do I start searching?
Thanks to all! Xavi.
-- Xavier Montero - 93 589 71 91 - 630 59 01 62 - xmontero@dsitelecom.com
I had the same problem due to clock backward
On 06/25/2010 09:53 AM, Xavi Montero wrote:
Hi all, my first post.
After years with an old server, a couple of months ago I've installed a new one.
We use IMAP. After research, I chosed dovecot.
It has been running for several weeks but from one week ago or so, it suddently dies about once per day.
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each night. We did this because "apparently" we did not a "soft time drift" installing the full ntp daemon, but I start thinking the time-jump crashes dovecot.
QUESTIONS:
- Might this time-jump cause dovecot to die?
- I don't know wehere the dovecot logs are written to. Where do I start searching?
Thanks to all! Xavi.
Xavi Montero wrote:
It has been running for several weeks but from one week ago or so, it suddently dies about once per day.
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each night. We did this because "apparently" we did not a "soft time drift" installing the full ntp daemon, but I start thinking the time-jump crashes dovecot.
QUESTIONS:
- Might this time-jump cause dovecot to die?
- I don't know wehere the dovecot logs are written to. Where do I start searching?
I would suggest running ntpdate once at startup and then have ntp run continuously.
Having step changes in time will likely cause other problems.
Better to treat the problem properly at the source rather than having to put sticking-plasters on everything.
Bill
Am 25.06.2010 10:35, schrieb William Blunn:
Xavi Montero wrote:
It has been running for several weeks but from one week ago or so, it suddently dies about once per day.
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each night. We did this because "apparently" we did not a "soft time drift" installing the full ntp daemon, but I start thinking the time-jump crashes dovecot. Exactly. On a production server you should alsways set the time on startup against a good source, make sure your timezone settings are correct, anbd then let ntpd keep the time precise using the same source .
QUESTIONS:
- Might this time-jump cause dovecot to die?
It WILL cause dovecot to die. You will see this in the logs once you found them.
- I don't know wehere the dovecot logs are written to. Where do I start searching? cf. http://wiki.dovecot.org/Logging
or wherever you configured it to go ....
- Might this time-jump cause dovecot to die?
- I don't know wehere the dovecot logs are written to. Where do I
I've confirmed in the logs that the die was really controlled and caused by the time-jump:
Here's the log:
" Jun 25 06:25:00 hidrogeno dovecot: Time just moved backwards by 9 seconds. This might cause a lot of problems, so I'll just kill myself now. http://wiki.dovecot.org/TimeMovedBackwards "
So I installed ntp to control time in a smooth way.
Thanks for confirming that timejumps were the most probable cause and helping me to find the logs to confirm that.
Thanks to all again. Xavi.
-- Xavier Montero - 93 589 71 91 - 630 59 01 62 - xmontero@dsitelecom.com
On 25.6.2010 10:53, Xavi Montero wrote:
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each
Time jumps, especially backwards, are not a good thing on a server. You should really consider changing from ntpdate to for example ntpd that takes care of keeping the clock synchronized, without unnecessarily stepping the clock. You should probably do this even if the actual cause of Dovecot crashing turns out not to be the time jumps.
Dovecot typically logs via syslog, so the logs go where your syslog writes them to. Usually something under /var/log.
Tapio
On 25/06/2010 08:53, Xavi Montero wrote:
The only change done in the server one week ago has been to install ntpdate, running once per day in cron-daily.
In the cron reports, ntpdate reports a time-jump of 9 seconds each night. We did this because "apparently" we did not a "soft time drift" installing the full ntp daemon, but I start thinking the time-jump crashes dovecot.
I have started to prefer chrony now that development has restarted on it (with one of the redhat guys leading the charge). At the end of the day it's just a time server, but it seems to be simpler to run and quite reliable
Regards
Ed W
participants (6)
-
Ed W
-
Frank Bonnet
-
Jakob Curdes
-
Tapio Sokura
-
William Blunn
-
Xavi Montero