Postfix - terminating on signal 15

[cryptic]

A few of my ISPConfig/CentOS 5 boxes have a Postfix hiccup every Sunday night for the past month.

I have this in my maillogs -

Oct 1 00:00:45 server postfix/postfix-script: stopping the Postfix mail system

Oct 1 00:00:53 server postfix/smtpd[19053]: warning: database /etc/postfix/virtusertable.db is older than source file etc/postfix/virtusertable

Oct 1 00:14:09 server postfix/smtpd[13746]: table hash:/etc/aliases(0,lock|fold_fix) has changed -- restarting

Oct 1 00:15:59 server postfix/smtpd[19053]: table hash:/etc/aliases(0,lock|fold_fix) has changed -- restarting

Oct 1 00:13:21 server postfix/master[28578]: terminating on signal 15

Oct 1 00:17:09 server postfix/postfix-script: fatal: the Postfix mail system is already running

Following this, Postfix is down until I start it.

[till]

Does this really only happen on every sunday? Then you should have a look at your weekly cronjobs if one of them affects postfix. ISPConfig has no cronjob that is run weekly, all cronjobs are at least run once a day.

[cryptic]

No, not every Sunday. It's happened 2 weeks in a row on a Sunday for one server, and just happened on another server on a Sunday.

[falko]

Maybe it's a good idea to install monit on your servers. It can restart Postfix if it finds it isn't running:

http://www.howtoforge.com/server_monitoring_with_munin_monit_debian_etch
http://www.tildeslash.com/monit/index.php

[bluethunder82]

To solve the error:

Oct 1 00:00:53 server postfix/smtpd[19053]: warning: database /etc/postfix/virtusertable.db is older than source file etc/postfix/virtusertable

It is my understanding to run the command:

postmap /etc/postfix/virtusertable

--
For the error:

Oct 1 00:14:09 server postfix/smtpd[13746]: table hash:/etc/aliases(0,lock|fold_fix) has changed -- restarting

Try running:
newaliases

The only time I've seen a similar error was went both the alias_maps and alias_database where not correct.

What is your output of:

postconf alias_maps alias_database

I have no idea about the Sunday issue.

Cheers

[cryptic]

What is your output of:

postconf alias_maps alias_database

I have no idea about the Sunday issue.

Cheers

alias_maps = hash:/etc/aliases
alias_database = hash:/etc/aliases

This issue happened again on Sunday.

I have tons of these errors in my maillog:

postfix/smtpd[11070]: warning: problem talking to server private/anvil: Connection timed out
postfix/smtpd[10060]: table hash:/etc/aliases(0,lock|fold_fix) has changed -- restarting

[till]

The question is which script or process is causing this as /etc/aliases is not managed by ISPConfig :confused:

[cryptic]

This is occurring a few times a week now.

It happened last night. Wierd thing is that my maillog.2 ends at Oct 28, maillog.1 has nov 4 from midnight to 4:30am, and then maillog starts at 4:30 am .

When postfix looks dead - If I try to run /etc/init.d/postfix stop it fails and I see postfix is still running. I have to kill this process before I can start postfix again.

I see a lot of this in my maillog,

postfix/postfix-script: fatal: the Postfix mail system is already running

[cryptic]

It looks like syslog AND ISPConfig are rotating a maillog. Could this be causing a conflict?

I see maillog, maillog.1, maillog.2, etc

and

maillog.31-10-07_23-59-01, etc etc

[till]

ISPConfig is just making a copy of the mail logfile to analyze it, it is not rotating the main logfile. But to be sure you can try to change the time of the ISPConfig mail log cronjob (in the root crontab) to run a few minues earlier or later.

[cryptic]

This issue is now happening daily. Any other ideas?

[till]

You can try to comment out the mal log script from ISPConfig in the root crontab file for one night and see if potfix does not crash then.

[cryptic]

I have done this last week. It made no difference.

[till]

Ok, then please check your syslog. Is there any cronjob started just before postfix dies?

[cryptic]

Only check_services.php, which runs every 30 minutes.

[till]

I'am getting out of ideas :confused:

[cryptic]

This happened again this morning. I see the anvil process died -

22815 ? Z 0:13 [anvil] <defunct>

[Martin1982]

I'm having exactly the same issue, anyone came up with a solution yet?