Problem with Parallel Processes

fdegir has asked for the wisdom of the Perl Monks concerning the following question:

Hi, I have a script which spawns 5 parallel processes. This script runs correctly when I use it but when another user tries to run it, it starts correctly but in couple of seconds, 2 processes exit and remaining 3 processes continue running as normal. What can be the reason for this? Addition: Below script is not the script itself. I just wanted to show the part processes created. In real script, each process takes couple of hours to finish. 2 exiting processes don't touch files in the script when different user tries to run it. Simplified script is below,

#!/usr/bin/env perl
for($i = 1; $i < 6; $i++){
   my $pid = fork();
   if($pid){
      print "INFO\t: Started job $i (PID: $pid)\n";
      push @childs, $pid;
   }elsif($pid == 0){
      pJob($i);
      exit 0;
   }else{
      die "ERROR\t: Can not fork:$!\n";
   }
}

for(@childs){
   my $tmp = waitpid($_, 0);
}

sub pJob($){
   my $id = shift;
   `touch /tmp/file_$id`;
   print "INFO\t: Done with job $i\n";
}
[download]

OS is solaris10. Any comments?

Comment on Problem with Parallel Processes Download Code

Replies are listed 'Best First'.
Re: Problem with Parallel Processes by almut (Canon) on May 31, 2010 at 14:40 UTC
but in couple of seconds, 2 processes exit and remaining 3 processes continue running As the given script doesn't even run a single second, I suppose this is not the real code that actually exhibits the problem... So, can you replicate the issue when you add a simple sleep to `pJob()`? If not, what is the real code doing that might make the two processes terminate prematurely?	[reply] [d/l]
Re: Problem with Parallel Processes by JavaFan (Canon) on May 31, 2010 at 14:43 UTC
First of all, you aren't correctly determining the fork fails. If the fork fails, $pid will be undefined. But `undef == 0` is a true statement (which will generate a warning, if you have them turned on). I don't understand the but in couple of seconds, 2 processes exit and remaining 3 processes continue running as normal bit. All the processes are doing are forking, printing, and touching files. I'd expect the entire process, including the children, to be finished within a single second. Note however that if another user runs this as well, the `touch` will likely fail (depending on the umask of the original user) - but that shouldn't matter in whether or not the processes fail (and hardly in how fast they are).	[reply] [d/l] [select]