• RunningDovecot

Running Dovecot

Starting

Dovecot can simply be started by running dovecot as root. If there are any problems, they're usually written to terminal, but they may also be written to error log as well.

If you wish to, you can also start Dovecot from inetd or xinetd, but don't try that before you know the Dovecot installation is actually working.

• A sample universal init.d script.

• A sample Mac OS X 10.4 launchd plist file

Stopping

Killing the Dovecot master process with a normal TERM signal does a clean shutdown:

# master.pid is written to directory pointed by base_dir setting.
# /var/run/dovecot/ is a common location for it.
base_dir=`dovecot -a | grep ^base_dir: | sed 's/^[^:]*: //'`
kill `cat $base_dir/master.pid`

shutdown_clients setting controls whether existing IMAP and POP3 sessions are killed.

Processes

When Dovecot is running, it uses several processes:

# ps auxw|grep "dovecot\|imap\|pop3"
root     25992  0.0  0.1   3512  1248 ?        Ss   Jan02   2:49 dovecot
nobody   25993  0.0  0.1   6260  1920 ?        S    Jan02   0:10 dovecot-auth
nobody   27570  0.0  0.1   6264  1928 ?        S    22:45   0:00 dovecot-auth -w
dovecot  25994  0.0  0.1   3276  1440 ?        S    Jan02   0:19 pop3-login
dovecot  16809  0.0  0.1   3284  1444 ?        S    04:23   0:03 imap-login
user     27572  0.0  0.0   2156   740 ?        S    22:45   0:00 imap
user     27734  0.0  0.0   2096   948 ?        S    22:51   0:00 pop3

• dovecot process is the Dovecot master process which keeps everything running.

• dovecot-auth process handles all authentication.

• dovecot-auth -w process is an "authentication worker" process. It's used only with some "blocking" authentication databases, such as MySQL.

• imap-login and pop3-login processes handle new IMAP and POP3 connections until user has logged in.

• imap and pop3 processes handle the IMAP and POP3 connections after user has logged in.

Reloading Configuration

Send HUP signal to dovecot process. An acknowledgement is written to log file:

dovecot: Jan 08 00:36:45 Warning: SIGHUP received - reloading configuration

Note that log file locations aren't changed unless you fully stop and restart Dovecot.

Running Multiple Invocations of Dovecot

You may wish to invoke a second session (or even multiple sessions) of Dovecot for testing different functionality, configurations, etc. In order to to run multiple instances of Dovecot, you must:

1. Create a differently named copy of the dovecot.conf configuration file with these changes:

   1. Change base_dir to the new run directory

   2. Make sure that login_dir is either commented out (default is relative to base_dir), or change it under the new directory as well

   3. Change the listen and ssl_listen port numbers to new, unused values

   4. If you're using authentication sockets (for SMTP AUTH or deliver), you'll need to change them as well. auth_socket_path specifies the socket path for deliver.

      • Alternatively if all the instances have identical authentication configuration, you can have only a single Dovecot instance serve the auth sockets and have the other instances use them. It is mean, you can skip part

           1 socket listen {
           2   master { ... }
           3   client { ... }
           4 }
        

        inside auth default.

   5. Invoke dovecot (and deliver) with the -c flag and the modified configuration file, e.g.: dovecot -c /usr/local/etc/dovecot2.conf

Rotating Log Files

If you specified log file paths manually in dovecot.conf instead of using syslog, you can send USR1 signal to Dovecot to make it close and reopen the log files.

Troubleshooting

If you can't see the Dovecot processes running after starting dovecot, something is most likely wrong in your dovecot.conf. Look at the error from Dovecot's log file. See Logging for how to find the log.

If you really can't find any error messages from any logs, try starting Dovecot with dovecot -F. If you see it crash like:

sh: segmentation fault (core dumped)  dovecot -F

Then it's a bug in Dovecot. Please report it with your configuration file.

If it simply quits without giving any error, then it wrote the error to a log file and you just didn't find it. Try specifying the log file manually and make sure you're really looking at the correct file.