http://www.perlmonks.org?node_id=985426


in reply to Concurrent Cache Pattern

pileofrogs,

Because of your concern about a 'flock' on the lock from a runaway process, you can use 2 files, one for the locking and one for the information your caching. I'm going to show untested code to give you an idea of how to go about it, but only you can decide what to do in a true deadlock situation.

my $lock_file = "./LockFile"; ## Put this in a common place GetLock: if ( ! -e $lock_file ) ## No Process is working { open ( my $lock, ">", $lock_file ) or die "Can't open $! "; if ( flock ( $lock, LOCK_EX | LOCK_NB ) ) { print $lock "$$\n"; close $lock; ## Now, you need to make sure your the process with the lock open ( $lock, "<", $lock_file ) or die "Can't open $! "; if ( flock ( $lock, LOCK_SH ) ) { my $pid = < $lock >; chomp ( $pid ); close $lock; if ( $pid eq "$$" ) { last; } ## Good--you have the lo +ck else { redo GetLock; } } else { redo GetLock; } } else { ## Another process has the exclusive lock close $lock; usleep 1_000; ## wait .001 seconds redo GetLock; } } elsif { if ( -s $lock_file ) ## Another Process is already w +orking { open ( my $lock, "<", $lock_file ) or die "Can't open $! " +; my $who = < $lock >; chomp ( $who ); close $lock; ## Now you can stat the file and if it exists longer than ## some number of seconds/minutes, you have the pid to k +ill it. } else { redo GetLock; } } ## Put your testing and cache here.

The idea is to verify you have the lock before starting the cache process. Once you are sure you have the lock, you can test the cache file to determine if it is fresh or stale. When you finish your work you 'unlink' the lock file.

Things to think about:

Good Luck!

"Well done is better than well said." - Benjamin Franklin