NAME
Lazy::Lockfile - File based locking for the lazy.
SYNOPSIS
use Lazy::Lockfile;
my $lockfile = Lazy::Lockfile->new() || die "Couldn't get lock!";
...
# Lock is released when $lockfile goes out of scope or your program exits.
DESCRIPTION
Lazy::Lockfile is a module designed for simple locking through the use
of lockfiles, requiring very little effort on the part of the developer.
Once the object is instanced, the lock will be held as long as object
exists. When the object is destroyed, the lock is released.
Locks are based around the existence of a named file, not around the use
of flock (though flock is used to synchronize access to the lock file).
Lazy::Lockfile is (usually) smart enough to detect stale lockfiles from
PIDs no longer running by placing the PID of the process holding the
lock inside the lockfile.
NOTES
Lazy::Lockfile is not safe for use on NFS volumes.
Lazy::Lockfile is not tested to interact correctly with other file
locking systems when used on the same lockfile.
Lazy::Lockfile uses kill (with signal zero) to determine if the lockfile
is stale. This works on most systems running as most users but there are
likely instances where this will fail. If this applies to your system,
you can use the no_pid option to disable the check.
If Lazy::Lockfile encounters a malformed lockfile (empty, containing
other text, etc), it will treat it as a corrupt file and overwrite it,
assuming the lock. The author believes this behavior should be changed
(and malformed files should be left untouched), but has kept this
behavior for backwards compatibility.
USAGE
All of the magic in Lazy::Lockfile is done through the constructor and
destructor.
METHODS
new
Constructor for Lazy::Lockfile.
Parameters
Accepts a single optional parameter, a hashref containing the following
options:
location
Specifies the full path to the location of the lockfile. Defaults to:
'/tmp/' . (fileparse($0))[0] . '.pid'
i.e., the name of the program being run, with a ".pid" extension, in
/tmp/.
no_pid
If true, instead of writing the PID file, a value of "0" is written
instead. When read by another instance of Lazy::Lockfile attempting to
acquire the lock, no PID check will be performed and the lock will be
assumed to be active as long as the file exists. Defaults to false.
delete_on_destroy
If true, sets the "delete on destroy" flag. This flag defaults to true,
which causes the lockfile to be removed when the object is destroyed.
Generally, this is the desired behavior. When set to false, this flag
prevents the lockfile from being removed automatically when the object
is destroyed. See also "delete_on_destroy".
Compatibility
For compatibility with older versions of Lazy::Lockfile (pre-1.0), a
single optional parameter is accepted, the path to the lockfile. This
parameter functions the same as the 'location' parameter described
above.
As stated above, malformed lockfiles will be overwritten, though this
may be subject to change in a future version.
Return value
If the lock can not be obtained, undef is returned (and $! will contain
useful information). Otherwise, the lock is exclusive to this process,
as long as the object exists.
Example
my $lockfile = Lazy::Lockfile->new( { location => "/var/lock", no_pid => 1 } )
|| die "Couldn't get lock!";
name
Returns the file name of the lockfile.
delete_on_destroy
Gets or sets the "delete on destroy" flag.
If called without a parameter (or with undef), delete_on_destroy will
return the current state of the "delete on destroy" flag. If called with
a parameter, this flag will be set.
unlock
Explicitly removes the lockfile, just as if the object were destroyed.
Once this has been called, delete_on_destroy will be set to false, since
the lock has already been deleted. Once this method is called, there is
not much use left for the object, so the user may as well delete it now.
unlock should be used when the lockfile needs to be removed
deterministically while the program is running. If you simply remove all
references to the Lazy::Lockfile object, the lock will be freed when
garbage collection is run, which is not guaranteed to happen until the
program exits (though it will likely happen immediately).
Returns a true value if the lockfile was found and removed, false
otherwise.
CHANGES
2017-05-29, 1.23 - jeagle
Detect and bail out on errors when writing to the lockfile (RT#121894).
Thanks MRDVT.
2014-10-30, 1.22 - jeagle
Add missing dependency.
2014-09-14, 1.21 - jeagle
Re-package to make it easier to convert to RPM, etc.
2012-04-01, 1.20 - jeagle
Updated documentation, thanks Alister W.
2011-01-05, 1.19 - jeagle
Change to unit tests to appease cpantesters.
2011-01-04, 1.18 - jeagle
Implement suggestion by srezic to check PIDs belonging to other users
(RT#69185).
Clean up documentation.
2010-06-22, 1.17 - jeagle
Update unlock to return a useful status.
2010-06-22, 1.16 - jeagle
Version bumps for migration to CPAN.
2009-12-03, 1.14 - jeagle
Fix a bug causing lockfiles with no_pid to not be deleted on
destroy/unlink.
2009-12-03, 1.13 - jeagle
Add the unlock method, to allow for deterministic lockfile removal at
runtime.
2009-11-30, 1.12 - jeagle
Update documentation to clarify delete_on_destroy parameter default
setting.
2009-07-06, 1.11 - jeagle
Fix error thrown when running with taint checking enabled.
2009-07-06, 1.10 - jeagle
Fix a bug with lockfile location being overwritten with the default.
2009-07-06, 1.9 - jeagle
Add new parameter, no_pid, which disabled active lockfile checks.
Allow constructor to accept multiple parameters via hashref.
2009-06-10, 0.4 - jeagle
Introduce the delete_on_destroy flag.
2009-06-03, 0.3 - jeagle
Open pid file with O_NOFOLLOW, to avoid symlink attacks.
Change default pid file location from /var/tmp to /tmp.
Correct dates in CHANGES section.
Add useful error indicators, documentation on error detection.
2009-04-27, 0.2 - jeagle
Fix a bug with unspecified lockfile paths trying to create impossible
file names.
2009-04-06, v0.1 - jeagle
Initial release.