Logging
Perforce Swarm is a web application, so there are several types of logging involved during the course of Swarm's normal operations.
Web server logging
All accesses to Swarm may be logged by the web server hosting Swarm. As web server log configuration is web server specific, refer to your web server's documentation. Since we recommend the use of Apache, more information regarding log configuration in Apache can be found here:
Perforce service logs
Swarm communicates with the associated Perforce service with every page request. Review the Swarm-generated requests on your Perforce service by enabling logging.
Tip
For more information, see:
Perforce System Administrator's Guide: Logging errors
Perforce System Administrator's Guide: Auditing user file access
Swarm logs
Depending on the log configuration you provide to Swarm, Swarm can log its own operations. Swarm's logs are much more helpful if you encounter problems.
Swarm stores its log data in the data/log
file. The
volume of entries recorded in the log depends on the configuration stored
in the data/config.php
file. Here is an example:
<?php
// this block should be a peer of 'p4'
'log' => array(
'priority' => 3, // 7 for max, defaults to 3
),
The maximum value for the log priority is 7
; higher
values do not result in increased logging. The minimum value is
0
, which means no logging; lower values do not result
in further logging reductions.
Trigger Token Errors
If the trigger tokens are missing or invalid, the web server error log contains a suitable error:
queue/add attempted with invalid/missing token: token used
A token is invalid when it is not formed from
the characters A
through Z
(in upper or lowercase), 0
through
9
, or a dash (-
).
A token is missing when a file using the token as its
name does not exist in the data/queue/tokens
directory.
Performance logging
Swarm logs warnings whenever commands issued to the Perforce service take longer than expected to complete. These warnings can be used to diagnose performance problems in the Perforce service.
Note
By default, warnings are not captured in the Swarm log. To capture
warnings, the log priority
must be set to 4
or higher.
The default configuration is:
<?php
// this block should be placed within the 'p4' block
'slow_command_logging' => array(
3, // commands without a specific rule have a 3-second limit
10 => array('print', 'shelve', 'submit', 'sync', 'unshelve'),
),
In this configuration block, the numeric key specifies the time threshold in seconds, and the value (if present) is an array of Perforce service commands that should use the threshold. Should a command be associated with multiple thresholds, the largest is used for that command.
In addition, Swarm automatically detects when the PHP extension
xhprof
is installed and collects profiling data for
requests that take longer than expected. The profiling data could be
helpful in diagnosing performance issues within Swarm itself, particularly
when analyzed in combination with the slow command logging described
above. When collected, profiling data is stored in the
data/xhprof
folder.
The default configuration is:
<?php
// this block should be a peer of 'p4'
'xhprof' => array(
'slow_time' => 3, // the threshold in seconds
'ignored_routes' => array('download', 'imagick', 'libreoffice', 'worker'),
),
slow_time
specifies the threshold, in seconds, that
should be used to determine when a Swarm request is
slow. ignored_routes
is an
array that specifies a list of Zend Framework route
names that should not be profiled. For example, Swarm's
Files
module specifies that the
download
route should be ignored from profiling as
downloads could take significantly longer than the default threshold.
Note
Depending on the performance of the server hosting Swarm, and
particularly the performance of the associated Perforce service, you may
want to monitor the data/xhprof
folder for disk
usage. Each request that exceeds the time threshold causes 200-600KB of
data to be written.