Re: Long running Perl Tk processes

[thundergnat]

I have a Perl::Tk based script that does state mandated process monitoring and data collection for some industrial processes in the manufacturing plant where I work. (Serial interface to some National Instrument FieldPoint IO.) We need to do continuous monitoring with some fairly substantial fines if we exceed a threshold of downtime so I was concerned about this myself.

I ended up setting up the script to re exec itself once a day to avoid any side effects of memory leaks. You lose a second or two while it restarts but memory leaks can't get big enough to cause problems.

Here are the relevant bits. (Note: this is chopped out of a much larger script so some variables are artificially localized and/or not used in this fragment.)

use warnings;
use strict;
use Tk;

use Date::Calc qw(
  Today_and_Now
  Add_Delta_DHMS
);

use constant OS_Win => $^O =~ /Win/;

my $last_time;
my $time_zone = -5;    # GMT-5 or, EST. Set TZ appropriately

my $mw = MainWindow->new;

my $repeat = $mw->repeat( 250, \&update );

MainLoop;

sub update{
    my ( $year, $month, $day, $hour, $minute, $second ) = 
      Add_Delta_DHMS( Today_and_Now( [localtime] ), 0, $time_zone, 0, 
+0 );
    
    # Restart the program every day at midnight. Sidestep a bunch of
    # memory leak problems.
    if ( ( "$hour$minute$second" eq '000' ) and ( $last_time eq '23595
+9' ) ) {
        OS_Win ? exec "wperl $0" : exec "perl $0 &";
    }
    $last_time = "$hour$minute$second";
    
    warn "$last_time\n"; # for testing purposes
}

I'm not saying this is the only or best solution, but it works for me. The script this was taken from has been running "continuously" for the past several years.