postfix is not running

[aapthorp]

I've just installed Zimbra on Debian 4.0 and have a problem getting postfix running. I've looked at the various threads and taken care of competing MTAs. Exim 4 was running, but is now deinstalled.

zmcontrol gives me

mta Stopped
postfix is not running
saslauthd is running with pid 12551
snmp Running

The zimbra log shows:

Oct 21 13:53:53 localhost postfix/postfix-script[12523]: warning: not owned by root: /opt/zimbra/postfix-2.4.3.3/conf/main.cf
Oct 21 13:53:53 localhost postfix/postfix-script[12537]: starting the Postfix mail system
...snip...
Oct 21 13:54:30 localhost postfix/postqueue[12976]: fatal: Queue report unavailable - mail system is down

sh -x bin/zmmtactl status indicates that it is indeed postfix that is failing.

If I try connecting to port 25 I get connection refused.

Any thoughts, as far as I can tell /etc/hosts and DNS is configured correctly

thanks

Adrian

[jeffw]

I'm seeing the same thing with Debian 4.0 and 5.0RC1.

Starting postfix by hand under strace appears to show it crashing:

--- SIGSEGV (Segmentation fault) @ 0 (0) ---
Process 18205 detached

[jeffw]

Here's the last bit of the output of strace:

24704 munmap(0x40020000, 4096) = 0
24704 open("/var/lib/misc/services.db", O_RDWR|O_LARGEFILE) = -1 ENOENT (No such file or directory)
24704 open("/var/lib/misc/services.db", O_RDWR|O_LARGEFILE) = -1 ENOENT (No such file or directory)
24704 open("/var/lib/misc/services.db", O_RDWR|O_LARGEFILE) = -1 ENOENT (No such file or directory)
24704 --- SIGSEGV (Segmentation fault) @ 0 (0) ---


I'm not sure if the failure to find /var/lib/misc/services.db is critical or not.

[jeffw]

I finally found the core file (in postfix/spool):

Core was generated by `/opt/zimbra/postfix-2.4.3.3/libexec/master'.
Program terminated with signal 11, Segmentation fault.
#0 0x080d5219 in __db_close_pp (dbp=0x811c1b8, flags=0) at ../dist/../db/db_iface.c:245
245 ../dist/../db/db_iface.c: No such file or directory.
in ../dist/../db/db_iface.c
(gdb) bt
#0 0x080d5219 in __db_close_pp (dbp=0x811c1b8, flags=0) at ../dist/../db/db_iface.c:245
#1 0x406a977f in ?? () from /usr/lib/libnss_db.so.2
#2 0x0811c1b8 in ?? ()
#3 0x00000000 in ?? ()

[jeffw]

seems to be the same problem described here:

MTA not starting, yet I'm unable to find much debug info...

[aapthorp]

Thanks for pointing me at the other post. Unfortunately the suggested fix doesn't appear to help. Did you have any luck?

[jholder]

Originally Posted by aapthorp

Thanks for pointing me at the other post. Unfortunately the suggested fix doesn't appear to help. Did you have any luck?

Can you please post your logs?

[aapthorp]

Problem solved. It seems that deinstalling exim4 with apt-get remove does not actually eradicate exim4!!!

So the sequence of events was something like this:

1) Problem with running postfix
2) Realise exim4 is still running. Stop exim4 and remove or so I thought
3) Problem as originally posted
4) Determine that the cause is the nsswitch bdb problem highlighted above. Reconfigure nsswitch.conf. Reboot.
5) Postfix still not running
6) Eventually spot that exim4 is still installed and of course restarted on reboot

Having done some hunting around it appears that exim4 is notoriously difficult to remove and the best recommendation is to disable it with:

update-rc.d -f exim4 remove

See Permanently getting rid of exim! | debianHELP

I think a note on this needs to be added to install guide.

[Dirk]

The sysv-rc-conf tool is great for this kind of thing. install with 'aptitude install sysv-rc-conf' and you're good to go.

This will allow easy control over what services start in each runlevel. Stopping exim in runlevel 2 will be sufficient as that's what Debian runs in as standard.

[dogatemycomputer]

I am posting because I tend to forget solutions later and I hope this helps the next person.

I changed the hostname for our server and rebooted. I thought this was the problem so I spent quite a bit of time trying to figure out why the hostname would prevent the MTA from starting. After some further investigation I found postfix had been completely uninstalled (which really means I did something stupid some time ago and the problem did not rear its ugly head until I unloaded the running postfix service.)

Here are the steps I went through on CentOS to fix the problem:

//realized postfix was not loaded by checking to see what was installed
rpm -qa | grep postfix

//installed postfix
yum install postfix

//postfix would load then exit because sendmail was already using port 25
chkconfig --level 345 sendmail off
chkconfig --level 345 postfix on
/etc/init.d/sendmail stop
/etc/init.d/postfix start

//confirmed postfix was loaded
telnet localhost 25


This all went well until 'su zimbra && zmcontrol status && "MTA not loaded"'. I knew for sure Postfix was loaded and working but for some reason Zimbra would not recognize this. I figured out that you have to stop Postfix, su zimbra then restart zimbra. If Zimbra loads postfix then everything will be fine but loading Postfix as root is not so good.

//do this as root
/etc/init.d/zimbra stop

//switch user to zimbra
su zimbra

//restart zimbra
zmcontrol restart

//wait patiently
zmcontrol status

//you should see the following:
mta Running


This obviously won't help everyone but understanding that Postfix cannot be loaded as root but needs to be loaded as user zimbra was interesting and fixed my problem.

Have a good night!