This documentation is for Dovecot v1.x, see wiki2 for v2.x documentation.
Differences between revisions 1 and 13 (spanning 12 versions)
Revision 1 as of 2006-01-07 20:42:00
Size: 1542
Editor: TimoSirainen
Comment:
Revision 13 as of 2007-08-16 15:14:14
Size: 3646
Editor: TimoSirainen
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
[[TableOfContents]]
Line 3: Line 5:
By default Dovecot starts logging to syslog using mail facility. In most operating systems this goes to `/var/log/mail.log`, `/var/log/maillog` or something similar. '''Dovecot always logs a detailed error message''' if something goes wrong. If it doesn't, it's considered a bug and will be fixed. However almost always the problem is that you're looking at the wrong log file. By default Dovecot logs to syslog using '''mail''' facility. You can change the facility from {{{syslog_facility}}} setting. You can also configure Dovecot to write to log files directly, see below.
Line 5: Line 7:
If there is anything wrong with your configuration or if something crashes, '''Dovecot will always log the error message'''. It's extremely rare that Dovecot dies without giving any sort of error message, so make sure you're looking at the correct log file. When using syslog, Dovecot uses 4 different logging levels:
Line 7: Line 9:
After you start Dovecot, make sure that you can find a line similar to this:  * '''info''': Informational and debug messages.
 * '''warning''': Warnings that don't cause an actual error, but are useful to know about.
 * '''err''': Non-fatal errors.
 * '''crit''': Fatal errors that cause the process to die.
Line 9: Line 14:
{{{
dovecot: Jan 07 14:22:40 Info: Dovecot v1.0.alpha5 starting up
}}}
Where exactly these messages are logged depends entirely on your syslog configuration. Often everything is logged to {{{/var/log/mail.log}}} or {{{/var/log/maillog}}}, and '''err''' and '''crit''' are logged to {{{/var/log/mail.err}}}. This is not necessarily true for your configuration though.

You can find the correct log files using these methods:

 * Info log: After starting Dovecot, {{{grep "starting up" /var/log/*}}}. It should show a line such as: {{{Dovecot v1.0.0 starting up}}}
 * Error logs: Use {{{dovecot --log-error}}} command, which makes Dovecot log some errors and exit. Then {{{grep "This is Dovecot's" /var/log/*}}} to find them. You should see:
  * With Dovecot v1.0.0 you'll find only the '''crit''' log: {{{This is Dovecot's error log}}}
  * With Dovecot v1.0.1 you'll find all of them:
   * '''warning''': {{{This is Dovecot's warning log}}}
   * '''err''': {{{This is Dovecot's error log}}}
   * '''crit''': {{{This is Dovecot's fatal log}}}
 * You can also check your {{{/etc/syslog.conf}}} to see how it's configured.
Line 14: Line 28:
Line 20: Line 33:
The point is that whenever anything unexpected happens, Dovecot doesn't leak any extra information about it to clients. They don't need it and they might try to exploit it in some ways, so the less they know the better.
Line 21: Line 35:
The real error message is written to Dovecot's log file. The timestamp is meant to help you find it. The real error message is written to the error log file. The timestamp is meant to help you find it.
Line 24: Line 38:

If you don't want to use syslog, or if you just can't find the Dovecot's logs, you can make Dovecot log elsewhere as well:
If you don't want to use syslog, or if you just can't find the Dovecot's error logs, you can make Dovecot log elsewhere as well:
Line 29: Line 42:
# If you want everything in one file, just don't specify log_info_path
log_info_path = /var/log/dovecot-info.log
# If you want everything in one file, just don't specify info_log_path
info_log_path = /var/log/dovecot-info.log
}}}
The error messages will go to file specified by {{{log_path}}}, while everything else goes to {{{info_log_path}}}. If you do this, make sure you're really looking at the {{{log_path}}} file for error messages, since the "Starting up" message is written to {{{info_log_path}}} file.

== Rotating Logs ==
If you change from syslog to an external log file, you can use logrotate (available on most recent linux distros) to maintain the Dovecot logfile so it doesn't grow beyond a manageable size. Save the below scriptlet as {{{/etc/logrotate.d/dovecot}}}:

{{{
# dovecot SIGUSR1: Re-opens the log files.
/var/log/dovecot*.log {
  missingok
  notifempty
  delaycompress
  sharedscripts
  postrotate
    /bin/kill -USR1 `cat /var/run/dovecot/master.pid 2>/dev/null` 2> /dev/null || true
  endscript
}
Line 33: Line 63:
The error messages will go to file specified by `log_path`, while everything else goes to `log_info_path`. If you do this, make sure you're really looking at the `log_path` file for error messages, since the "Starting up" message is written to `log_info_path` file. '''NOTE''': change the path to the logfile(s) and the master.pid file as appropriate for your system configuration.

TableOfContents

Dovecot Logging

Dovecot always logs a detailed error message if something goes wrong. If it doesn't, it's considered a bug and will be fixed. However almost always the problem is that you're looking at the wrong log file. By default Dovecot logs to syslog using mail facility. You can change the facility from syslog_facility setting. You can also configure Dovecot to write to log files directly, see below.

When using syslog, Dovecot uses 4 different logging levels:

  • info: Informational and debug messages.

  • warning: Warnings that don't cause an actual error, but are useful to know about.

  • err: Non-fatal errors.

  • crit: Fatal errors that cause the process to die.

Where exactly these messages are logged depends entirely on your syslog configuration. Often everything is logged to /var/log/mail.log or /var/log/maillog, and err and crit are logged to /var/log/mail.err. This is not necessarily true for your configuration though.

You can find the correct log files using these methods:

  • Info log: After starting Dovecot, grep "starting up" /var/log/*. It should show a line such as: Dovecot v1.0.0 starting up

  • Error logs: Use dovecot --log-error command, which makes Dovecot log some errors and exit. Then grep "This is Dovecot's" /var/log/* to find them. You should see:

    • With Dovecot v1.0.0 you'll find only the crit log: This is Dovecot's error log

    • With Dovecot v1.0.1 you'll find all of them:
      • warning: This is Dovecot's warning log

      • err: This is Dovecot's error log

      • crit: This is Dovecot's fatal log

  • You can also check your /etc/syslog.conf to see how it's configured.

Internal Errors

If IMAP or POP3 processes encounter some error, they don't show the exact reason for clients. Instead they show:

Internal error occurred. Refer to server log for more information. [2006-01-07 22:35:11]

The point is that whenever anything unexpected happens, Dovecot doesn't leak any extra information about it to clients. They don't need it and they might try to exploit it in some ways, so the less they know the better.

The real error message is written to the error log file. The timestamp is meant to help you find it.

Changing Log File Paths

If you don't want to use syslog, or if you just can't find the Dovecot's error logs, you can make Dovecot log elsewhere as well:

log_path = /var/log/dovecot.log
# If you want everything in one file, just don't specify info_log_path
info_log_path = /var/log/dovecot-info.log

The error messages will go to file specified by log_path, while everything else goes to info_log_path. If you do this, make sure you're really looking at the log_path file for error messages, since the "Starting up" message is written to info_log_path file.

Rotating Logs

If you change from syslog to an external log file, you can use logrotate (available on most recent linux distros) to maintain the Dovecot logfile so it doesn't grow beyond a manageable size. Save the below scriptlet as /etc/logrotate.d/dovecot:

# dovecot SIGUSR1: Re-opens the log files.
/var/log/dovecot*.log {
  missingok
  notifempty
  delaycompress
  sharedscripts
  postrotate
    /bin/kill -USR1 `cat /var/run/dovecot/master.pid 2>/dev/null` 2> /dev/null || true
  endscript
}

NOTE: change the path to the logfile(s) and the master.pid file as appropriate for your system configuration.

None: Logging (last edited 2009-10-28 03:05:44 by TimoSirainen)