In its simplest form, lumberjack is a pipe logger for Apache httpd and other daemons.
Lumberjack can be used as the logger for piped log files using the httpd CustomLog directive - processing each log line and writing it out to the correct - per VirtualHost - log file. It may also be used by other daemons capable of writing log files to a pipe/FIFO.
In contrast to having one logger process per VirtualHost (or a hard coded log file for each VirtualHost, as is the norm), lumberjack can be set as the single pipe logger in the global section of the httpd.conf file, and will - with an appropriate LogFormat specifier - split off each log line for a different VirtualHost and write it to a per VirtualHost log file based upon a user defined template.
Lumberjack may also be used in 'raw' mode in combination with the httpd ErrorLog directive to log errors to a log file based upon a user supplied template, either on a per server or per VirtualHost basis. 'raw' mode also allows other daemons to log via lumberjack, either as a pipe logger or via a FIFO.
- Automatic log file compression (using a user specifiable compressor) after log files are rotated.
- Request flushing/syncing of log files after each write (though this is not recommended).
- Reading of log lines from a FIFO rather than stdin - allowing use with other daemons such as ftpd and rsyncd.
- Automatic creation of a symbolic link to the currently active log file.
- 'raw' logging mode, for use with HTTPd's ErrorLog and with other daemons such as ftpd and rsyncd.
- User specified umask settings for directories and files.
- Log file names based upon a strftime() encoded template, optionally including the HTTPd VirtualHost identifier.
lumberjack [options] <basedir> <template>
There are two mandatory arguments when using lumberjack; the base directory of the log file path (<basedir>) and the template of the log file path (<template>). All other [options] are not required and only serve to modify the default behaviour of lumberjack.
The base directory (<basedir>) is the path to the root of where the log files should be written. This could be, for example, the root of the tree of sites which are virtual hosted, each with their own logs/ sub-directory. Or it could be a seperate directory tree specifically for log files. The <basedir> is pretty flexible and highly dependant upon your local filesystem layout.
Personally, I use a layout resembling:
/data/sites/ - This would be my `<basedir>`.
<virtualhost name>/ - This is the name of the site served by the `VirtualHost`.
cgi-bin/ - The virtual host's CGI binaries directory.
html/ - The virtual host's web content.
logs/ - The logs directory for the virtual host.
Under the logs/ directory, I have files written in the format: YYYY/MM/filename.log - where YYYY is the year, and MM is the two digit month. An example full logfile path would therefore be: /data/sites/afterdark.org.uk/logs/2018/04/httpd-access.log.
To accomplish this format for a log file, the <template> would need to be set to: {}/logs/%Y/%m/httpd-access.log. The sequence {} appearing anywhere (even multiple times) in the template will be replaced with the site name as taken from VirtualHost site identifier from the httpd.conf (see usage below for how to configure this for use with lumberjack). The %Y and %m are strftime() encoded expansions which would yeild the current year, and two digit month.
The templaing feature of lumberjack is one of its most powerful features. It can be used to create any free form log file path under the <basedir>.
A full set of currently supported options can be obtained using lumberjack -h - this will always show the most up to date usage and options.
The usages detailed below are the simplest form of the logger - options such as -ca, -cc, -f, -j, -md and -mf which modify the behaviour of lumberjack will not be included in the examples. Using those options should be fairly self-explanitory after reading lumberjack -h output.
In this mode of operation, you do not need to configure a CustomLog for each VirtualHost you use in your configuration. Instead, the CustomLog directive is placed in the global section of the httpd.conf and will become effective for every VirtualHost (and the main server configuration, if used).
This is the recommended mode of operation for Apache httpd, as it only uses a single lumberjack process to handle logs for every VirtualHost defined in the configuration.
Before configuring lumberjack as the CustomLog pipe logger in httpd, it is necessary to add a custom LogFormat specifier for use with lumberjack. This new LogFormat specifier is the same as the standard Apache 'Common' or 'Combined' log formats - just with the addition of the '%v' specifier. lumberjack may be used with any set of LogFormat specifiers, as long as the first item in the specifier is always %v.
For the purposes of this usage example, we will be using the 'Combined' log format.
Create a new LogFormat entry (in this case named "VHostCombined") in the global section of httpd.conf, based upon the "Combined" format already given as an example in the default httpd.conf; but with the addition of the %v specifier for lumberjack:
LogFormat "%v %h %l %u %t \"%r\" %>s %b \"%{Referer}i\" \"%{User-Agent}i\"" VHostCombined
Then locate the CustomLog directive in the httpd.conf. Or you may need to add this if there is none already present.
The format for the CustomLog directive for use with lumberjack is:
CustomLog "|/path/to/lumberjack -l {}/logs/httpd-access.log -z /path/to/basedir {}/logs/%Y/%m/httpd-access.log" VHostCombined
Here, the -l and -z flags are extra options to lumberjack - they can be omitted if you do not require them. The template is just an example based upon my personal layout for the directories that was given as an example earlier; you can use any layout that suits your personal directory layout.
Lumberjack can also be used to handle error logs generated by Apache httpd in almost the same way as the CustomLog example above. The only difference is the addition of the -r option, to tell Lumberjack it should be working in 'raw' mode, and the LogFormat changes are not necessary.
Unlike in 'normal' mode, where lumberjack extracts the VirtualHost site identifier for use in the template (and removes that identifier from the lon entry written); in 'raw' mode, no procesing of the log line is performed - the VirtualHost identifier is not extracted to be used as part of the log file name, and the log line is written verbatim.
In 'raw' mode, the template (and link name if specified) cannot include the special {} string, which is usually replaced with the VirtualHost site identifier when in 'normal' mode - Lumberjack will not start if either of those options include a {} sequence.
Unfortunately, because Apache httpd does not allow custom formats for the error log (where we could include the site idetifier and only use one ErrorLog directive), we must use an ErrorLog per VirtualHost. An example ErrorLog directive, based upon the examples given previously, would be:
ErrorLog "|/path/to/lumberjack -l httpd-error.log -r -z /data/sites/afterdark.org.uk/logs %Y/%m-httpd-error.log"
Here, the <basedir> includes the full path to the logs directory to be used, since it cannot be parsed from the log line itself. The <template> and link name (-l) have also been changed to fit with the new <basedir>.
As a side benefit of the 'raw' logging mode which can be used for the ErrorLog handling above, lumberjack can be used as a generic pipe logger for any daemon which supports logging via a pipe or a FIFO (see next section for details on FIFO logging).
As long as a daemon can log via a pipe, lumberjack can be used in the same was as demonstrated in the ErrorLog section above. The only details that would need to be changed are the <basedir>, <template> and any configured link name.
Personally, I use lumberjack to handle the logging (and rotation/compression of old logs) for not only httpd, but also for ProFTPd and rsyncd - the latter two by making use of a FIFO.
Using lumberjack with daemons that do not support pipe logging is as simple as setting up a FIFO for them to write their logs to, and telling lumberjack where to find it.
To do this, use the -i flag to tell lumberjack where to locate the FIFO for the daemon from which you want to log. All other command line options are the same as detailed in the other examples, but it is most likely that you'll also want to use the -r flag, to tell lumberjack not to process the log lines for HTTPd VirtualHost site identifiers.