NAME
IPC::ConcurrencyLimit::WithStandby - IPC::ConcurrencyLimit with an additional standby lock
SYNOPSIS
use IPC::ConcurrencyLimit::WithStandby;
sub run {
my $limit = IPC::ConcurrencyLimit::WithStandby->new(
type => 'Flock', # that's also the default
max_procs => 10,
path => '/var/run/myapp',
standby_path => '/var/run/myapp/standby',
standby_max_procs => 3,
);
my $id = $limit->get_lock;
if (not $id) {
warn "Got none of the worker locks. Exiting.";
exit(0);
}
else {
# Got one of the worker locks (ie. number $id)
do_work();
}
# lock released with $limit going out of scope here
}
run();
exit();
DESCRIPTION
This module provides the same interface as the regular IPC::ConcurrencyLimit module. It differs in what happens if get_lock
fails to get a slot for the main limit:
If it fails to get a (or the) lock on the main limit, it will repeatedly attempt to get the main lock until a slot for the main limit is attained or the number of retries is exhausted. Most importantly, this supports limiting the number of instances that continuously attempt to get the main lock (typically, this would be limited to 1). This is implemented with a wait-retry-loop and two separate IPC::ConcurrencyLimit
objects.
The options for the main limit are passed in to the constructor as usual. The standby limit are inherited from the main one, but all parameters prefixed with standby_
will override the respective inherited parameters. For example, standby_type => "MySQL"
will enforce the use of the MySQL lock for the standby lock.
In addition to the regular IPC::ConcurrencyLimit
options, the constructor accepts retries
as the number of retries a standby instance should do to get the main lock. There will always be only one attempt to become a standby process. Additionally, interval
can indicate a number of seconds to wait between retries (also supports fractional seconds down to what Time::HiRes::sleep
supports).
retries
can also be passed a code reference that will be called on every retry, with the current attempt number as its first argument (starting at 1). Returning false from this routine will break the loop and give up any further attempts to acquire the lock. One example of a configuration which would continue to attempt to acquire a lock forever would be as follows:
my $limit = IPC::ConcurrencyLimit::WithStandby->new(
retries => sub { $_[0] },
interval => 0.01,
maxproc => 1,
standby_max_procs => 1,
...
);
The form above would be used to have a single process running, with a standby process ready to take over 1/100th of a second after the active process exits, at the expense of attempting to acquire a flock()
100 times per second.
Finally, as a way to tell blocked worker processes apart from standby processes, the module supports the process_name_change
option. If set to true, then the module will modify the process name of standby processes via modification of <$0>. It appends the string " - standby" to $0 and resets it to the old value after timing out or getting a worker lock. This is only supported on newer Perls and might not work on all operating systems. On my testing Linux, a process that showed as perl foo.pl
in the process table before using this feature was shown as foo.pl - standby
while in standby mode and as foo.pl
after getting a main worker lock. Note the curiously stripped perl
prefix.
AUTHOR
Steffen Mueller, smueller@cpan.org
David Morel, david.morel@amakuru.net
ACKNOWLEDGMENT
This module was originally developed for booking.com. With approval from booking.com, this module was generalized and put on CPAN, for which the authors would like to express their gratitude.
COPYRIGHT AND LICENSE
(C) 2012 Steffen Mueller. All rights reserved.
This code is available under the same license as Perl version
5.8.1 or higher.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.