Log::Log4perl::Appender::Synchronized - Synchronizing other appenders
use Log::Log4perl qw(:easy);
my $conf = qq( log4perl.category = WARN, Syncer # File appender (unsynchronized) log4perl.appender.Logfile = Log::Log4perl::Appender::File log4perl.appender.Logfile.autoflush = 1 log4perl.appender.Logfile.filename = test.log log4perl.appender.Logfile.mode = truncate log4perl.appender.Logfile.layout = SimpleLayout # Synchronizing appender, using the file appender above log4perl.appender.Syncer = Log::Log4perl::Appender::Synchronized log4perl.appender.Syncer.appender = Logfile );
Log::Log4perl->init(\$conf); WARN("This message is guaranteed to be complete.");
If multiple processes are using the same Log::Log4perl
appender
without synchronization, overwrites might happen. A typical scenario
for this would be a process spawning children, each of which inherits
the parent's Log::Log4perl configuration.
In most cases, you won't need an external synchronisation tool like
Log::Log4perl::Appender::Synchronized at all. Log4perl's file appender,
Log::Log4perl::Appender::File, for example, provides the syswrite
mechanism for making sure that even long log lines won't interleave.
Short log lines won't interleave anyway, because the operating system
makes sure the line gets written before a task switch occurs.
In cases where you need additional synchronization, however, you can use
Log::Log4perl::Appender::Synchronized
as a gateway between your
loggers and your appenders. An appender itself,
Log::Log4perl::Appender::Synchronized
just takes two additional
arguments:
appender
Log::Log4perl::Appender::Synchronized
.
key
Log::Log4perl::Appender::Synchronized
uses internally to ensure
atomic operations. It defaults to _l4p
. If you define more than
one Log::Log4perl::Appender::Synchronized
appender, it is
important to specify different keys for them, as otherwise every
new Log::Log4perl::Appender::Synchronized
appender will nuke
previously defined semaphores. The maximum key length is four
characters, longer keys will be truncated to 4 characters --
mylongkey1
and mylongkey2
are interpreted to be the same:
mylo
(thanks to David Viner <dviner@yahoo-inc.com> for
pointing this out).
Log::Log4perl::Appender::Synchronized
uses Log::Log4perl::Util::Semaphore
internally to perform locking with semaphores provided by the
operating system used.
The Log::Log4perl::Appender::Synchronized
serializes access to a
protected resource globally, slowing down actions otherwise performed in
parallel.
Unless specified otherwise, all instances of
Log::Log4perl::Appender::Synchronized
objects in the system will
use the same global IPC key _l4p
.
To control access to different appender instances, it often makes sense to define different keys for different synchronizing appenders. In this way, Log::Log4perl serializes access to each appender instance separately:
log4perl.category = WARN, Syncer1, Syncer2 # File appender 1 (unsynchronized) log4perl.appender.Logfile1 = Log::Log4perl::Appender::File log4perl.appender.Logfile1.filename = test1.log log4perl.appender.Logfile1.layout = SimpleLayout # File appender 2 (unsynchronized) log4perl.appender.Logfile2 = Log::Log4perl::Appender::File log4perl.appender.Logfile2.filename = test2.log log4perl.appender.Logfile2.layout = SimpleLayout # Synchronizing appender, using the file appender above log4perl.appender.Syncer1 = Log::Log4perl::Appender::Synchronized log4perl.appender.Syncer1.appender = Logfile1 log4perl.appender.Syncer1.key = l4p1
# Synchronizing appender, using the file appender above log4perl.appender.Syncer2 = Log::Log4perl::Appender::Synchronized log4perl.appender.Syncer2.appender = Logfile2 log4perl.appender.Syncer2.key = l4p2
Without the .key = l4p1
and .key = l4p2
lines, both Synchronized
appenders would be using the default _l4p
key, causing unnecessary
serialization of output written to different files.
To configure the underlying Log::Log4perl::Util::Semaphore module in a different way than with the default settings provided by Log::Log4perl::Appender::Synchronized, use the options parameter:
log4perl.appender.Syncer1.destroy = 1 log4perl.appender.Syncer1.mode = sub { 0775 } log4perl.appender.Syncer1.uid = hugo log4perl.appender.Syncer1.gid = 100
Valid keys are
destroy
(Remove the semaphore on exit),
mode
(permissions on the semaphore),
uid
(uid or user name the semaphore is owned by),
and
gid
(group id the semaphore is owned by),
Note that mode
is usually given in octal and therefore needs to be
specified as a perl sub {}, unless you want to calculate what 0755 means
in decimal.
Changing ownership or group settings for a semaphore will obviously only
work if the current user ID owns the semaphore already or if the current
user is root
.
Setting user and group IDs is especially important when the Synchronized appender is used with mod_perl. If Log4perl gets initialized by a startup handler, which runs as root, and not as the user who will later use the semaphore, the settings for uid, gid, and mode can help establish matching semaphore ownership and access rights.
Log::Log4perl::Appender::Synchronized
is a composite appender.
Unlike other appenders, it doesn't log any messages, it just
passes them on to its attached sub-appender.
For this reason, it doesn't need a layout (contrary to regular appenders).
If it defines none, messages are passed on unaltered.
Custom filters are also applied to the composite appender only. They are not applied to the sub-appender. Same applies to appender thresholds. This behaviour might change in the future.
Copyright 2003 by Mike Schilli, all rights reserved. This program is free software, you can redistribute it and/or modify it under the same terms as Perl itself.
2003, Mike Schilli <m@perlmeister.com>