#!/usr/bin/perl
#
# 2011/11/27 gabriel

use strict;

use Getopt::Long();
use Pod::Usage;
use Coro;
use Coro::Semaphore;
use Coro::Signal;
use Coro::Channel;
use Coro::Handle;
use IO::File;
use POSIX qw( WNOHANG WEXITSTATUS );
use Cwd qw( getcwd );

my $filecmd  = '';
my $dir      = '';
my $cmd      = '';
my $logtrace = '';
my $verbose;
my $job_np = 1;
my $nodefile = $ENV{OAR_NODE_FILE} || '';
my $masterio;
my $switchio;
my $help;
my $oarsh = 'oarsh -q -T';

Getopt::Long::GetOptions(
   'filecmd=s'  => \$filecmd,
   'dir=s'      => \$dir,
   'cmd=s'      => \$cmd,
   'logtrace=s' => \$logtrace,
   'verbose'    => \$verbose,
   'help'       => \$help,
   'oarsh=s'    => \$oarsh,
   'jobnp=i'    => \$job_np,
   'nodefile=s' => \$nodefile,
   'masterio=s' => \$masterio,
   'switchio'   => \$switchio,
   ) || pod2usage(-verbose => 0);
pod2usage(-verbose => 2) if $help;
pod2usage(-verbose => 2) if not (
 (-e "$filecmd")
 or (-d "$dir" and $cmd ne '')
 );

# re-run, keep trace of job already done
my %state;
my $log_h = IO::File->new();
if (-e "$logtrace") {
   $log_h->open("< $logtrace")
      or die "error: can't read log file: $!";
   while (<$log_h>) {
      $state{$1} = 'start' if m/^start\s+job\s+([^\s]+)\s/;
      $state{$1} = 'end'   if m/^end\s+job\s+([^\s]+)\s/;
      }
   $log_h->close();
   }
if ($logtrace) {
   $log_h->open(">> $logtrace")
      or die "error: can't append log file $logtrace: $!";
   $log_h->autoflush;
   $log_h = unblock $log_h;
   }

# job to run
my @job = ();
if (-e "$filecmd") {
   my $job_num = 0;
   open(JOB_LIST, '<', "$filecmd") or die "error: can't open job file $filecmd: $!";
   while (<JOB_LIST>) {
      chomp;
      next if m/^#/;
      next if m/^\s*$/;
      $job_num++;
      push @job, { name => $job_num, cmd => "$_" };
      }
   close JOB_LIST;
   }
else {
   opendir(DIR, $dir) or die "error: can't open folder $dir: $!";
   while (my $item = readdir(DIR)) {
      next if $item =~ m/^\./;
      next if $item =~ m/:/;
      next if $item =~ m/\.old$/;
      next if $item =~ m/\.sav$/;
      next if $item =~ m/\.bak$/;
      next if $item =~ m/\.no$/;
      next unless (-d "$dir/$item");
      push @job, { name => $item, cmd => "( cd $dir/$item/; $cmd )" };
      }
   closedir DIR;
   }

# ressources available
my @ressources = ();
open(NODE_FILE, '<', "$nodefile")
   or die "can't open $nodefile: $!";
while (<NODE_FILE>) {
   chomp;
   next if m/^#/;
   next if m/^\s*$/;
   push @ressources, $_;
   }
close NODE_FILE;

my $ressource_size = scalar(@ressources);
die "error: not enought ressources jobnp $job_np > ressources $ressource_size"
   if $job_np > $ressource_size;

my $current_dir = getcwd();

my $stderr = $ENV{OAR_STDERR} || '';
$stderr =~ s/\.stderr$//;
$stderr = $masterio if $masterio;
my $stdout = $ENV{OAR_STDOUT} || '';
$stdout =~ s/\.stdout$//;
$stdout = $masterio if $masterio;

my $finished = new Coro::Signal;
my $job_todo = new Coro::Semaphore 0;
my $job_name_maxlen;
for (@job) {
   $job_todo->up;
   $job_name_maxlen = length($_->{name}) if length($_->{name}) > $job_name_maxlen;
   }

# slice of ressources for parallel job
my $ressources = new Coro::Channel;
for my $slot (1 .. int($ressource_size / $job_np)) {
   $ressources->put(
      join(',',
         @ressources[ (($slot - 1) * $job_np) .. (($slot * $job_np) - 1) ])
         );
   }

my %scheduled = ();

# OAR checkpoint and default signal SIGUSR2
my $oar_checkpoint = new Coro::Semaphore 0;
$SIG{USR2} = sub {
   print "warning: receive checkpoint at "
      . time
      . ", no new job, just finishing running job\n"
      if $verbose;
   $oar_checkpoint->up();
   };

# asynchrone start job block
async {
	JOB:
   for my $job (@job) {
      my $job_name = $job->{name};
      my $job_cmd  = $job->{cmd};

      # job has been already run ?
      if (exists $state{$job_name}) {
         if ($state{$job_name} eq 'start') {
            print "warning: job $job_name was not clearly finished, relaunching...\n"
               if $verbose;
            }
         elsif ($state{$job_name} eq 'end') {
            delete $state{$job_name}; # free memory
            $job_todo->down;
            print "warning: job $job_name already run\n" if $verbose;
            cede;
            next JOB;
            }
         }

      # take job ressource
      my $job_ressource = $ressources->get;

      # no more launch job when OAR checkpointing
      last JOB if $oar_checkpoint->count() > 0;

      my ($node_connect) = split ',', $job_ressource;
      my $fh = IO::File->new();
      my $job_pid = $fh->open("| $oarsh $node_connect >/dev/null 2>&1")
         or die "error: can't start subjob: $!";

      $fh->autoflush;
      $fh = unblock $fh;

      $scheduled{$job_pid} = {
         fh           => $fh,
         node_connect => $node_connect,
         ressource    => $job_ressource,
         name         => $job_name
         };

      my $msg = sprintf "start job %${job_name_maxlen}s / %5i at %s on node %s\n",
         $job_name, $job_pid, time, $job_ressource;
      $log_h->print($msg) if $logtrace;
      print($msg) if $verbose;

      my ($job_stdout, $job_stderr);
      $job_stdout = ">  $stdout-$job_name.stdout" if $stdout ne '' and $switchio;
      $job_stderr = "2> $stderr-$job_name.stderr" if $stderr ne '' and $switchio;

      my $job_nodefile = "/tmp/oar-parexec-$ENV{LOGNAME}-$job_name";

     # set job environment, run it and clean
      if ($job_np > 1) {
         $fh->print("printf \""
               . join('\n', split(',', $job_ressource,))
               . "\" > $job_nodefile\n");
         $fh->print("OAR_NODE_FILE=$job_nodefile\n");
         $fh->print("OAR_NP=$job_np\n");
         $fh->print("export OAR_NODE_FILE\n");
         $fh->print("export OAR_NP\n");
         $fh->print("unset OAR_MSG_NODEFILE\n");
         }
      $fh->print("cd $current_dir\n");
      $fh->print("$job_cmd $job_stdout $job_stderr\n");
      $fh->print("rm -f $job_nodefile\n") if $job_np > 1;
      $fh->print("exit\n");
      cede;
      }
   }

# asynchrone end job block
async {
   while () {
      for my $job_pid (keys %scheduled) {
			# non blocking PID test
         if (waitpid($job_pid, WNOHANG)) {
            my $msg = sprintf "end   job %${job_name_maxlen}s / %5i at %s on node %s\n",
               $scheduled{$job_pid}->{name},
               $job_pid, time, $scheduled{$job_pid}->{ressource};
            $log_h->print($msg) if $logtrace;
            print($msg) if $verbose;
            close $scheduled{$job_pid}->{fh};
            # leave ressources for another job
            $ressources->put($scheduled{$job_pid}->{ressource});
            $job_todo->down;
            delete $scheduled{$job_pid};
            }
         cede;
         }

      # checkpointing ! just finishing running job and quit
      $finished->send if $oar_checkpoint->count() > 0 and scalar(keys(%scheduled)) == 0;

      $finished->send if $job_todo->count() == 0;
      cede;
      }
   }

cede;

# all job have been done
$finished->wait;

# close log trace file
$log_h->close() if $logtrace;

__END__

=head1 NAME

oar-parexec - parallel execution of many small job

=head1 SYNOPSIS

 oar-parexec --filecmd filecommand [--logtrace tracefile] [--verbose] \
            [--jobnp integer] [--nodefile filenode] [--masterio basefileio] [--switchio] [--oarsh sssh]
 oar-parexec --dir foldertoitemize --cmd commandtolaunch [--logtrace tracefile] [--verbose] \
            [--jobnp integer] [--nodefile filenode] [--masterio basefileio] [--switchio] [--oarsh sssh]
 oar-parexec --help

=head1 DESCRIPTION

C<oar-parexec> can execute lot of small job in parallel inside a cluster.
Number of parallel job at one time cannot exceed the number of core define in the node file
C<oar-parexec> is easier to use inside an OAR job environment
which define automatically these strategics parameters...
However, it can be used outside OAR.

Option C<--filecmd> or C<--dir> and C<--cmd> are the only mandatory parameters.

Small job will be launch in the same folder as the master job.
Two environment variable are defined for each small job
and only in case of parallel small job (option C<--jobnp> > 1).

 OAR_NODE_FILE - file that list node for parallel computing
 OAR_NP        - number of processor affected

The file define by OAR_NODE_FILE is created  in /tmp
on the node before launching the small job
and this file will be delete after job complete.
C<oar-parexec> is a simple script,
OAR_NODE_FILE will not be deleted in case of crash of the master job.

OAR define other variable that are equivalent to OAR_NODE_FILE:
OAR_NODEFILE, OAR_FILE_NODES, OAR_RESOURCE_FILE...
You can use in your script the OAR original file ressources
by using these variable if you need it.
 

=head1 OPTIONS

=over 12

=item B<-f|--filecmd filecommand>

File name which content job list.
For the JOB_NAME definition,
the first valid job in the list will have the number 1 and so on...

=item B<-d|--dir foldertoitemize>

Command C<--cmd> will be launch in all sub-folder of this master folder.
Files in this folder will be ignored.
Sub-folder name which begin with '.'
or finish with '.old', '.sav', '.bak', '.no' will either be ignored...

The JOB_NAME is simply the Sub-folder name.

=item B<-c|--cmd commandtolaunch>

Command (and argument to it) tha will be launch in all sub-folder
parameter folfer C<--dir>

=item B<-l|--logtrace tracefile>

File which log and trace running job.
In case of running the same master command (after crash for example),
only job that are not mark as done will be run again.
Be careful, job mark as running (start but not finish) will be run again.
Tracing is base on the JOB_NAME between multiple run.

This option is very usefull in case of crash
but also for checkpointing and idempotent OAR job.

=item B<-v|--verbose>

=item B<-j|--jobnp integer>

Number of processor to allocated for each small job.
1 by default.

=item B<-n|--nodefile filenode>

File name that list all the node where job could be launch.
By defaut, it's define automatically by OAR via
environment variable C<OAR_NODE_FILE>.

For example, if you want to use 6 core on your cluster node,
you need to put 6 times the hostname node in this file,
one per line...
It's a very common file in MPI process !

=item B<-m|--masterio basefileio> 

The C<basefileio> will be use in place of environment variable
C<OAR_STDOUT> and C<OAR_STDERR> (without extension) to build the base name of the small job standart output
(only use when option C<swithio> is activated).

=item B<-s|--switchio> 

Each small job will have it's own output STDOUT and STDERR
base on master OAR job with C<JOB_NAME> inside
(or base on C<basefileio> if option C<masterio>).
Example :

 OAR.151524.stdout -> OAR.151524-JOB_NAME.stdout

where 151524 here is the master C<OAR_JOB_ID>
and C<JOB_NAME> is the small job name.

=item B<-o|-oarsh command>

Command use to launch a shell on a node.
By default

 oarsh -q -T

Change it to C<ssh> if you are not using an OAR cluster...

=item B<-h|--help>

=back


=head1 EXAMPLE

=head2 Simple list of sequential job

Content for the job file command (option C<--filecmd>) could have:

 - empty line
 - comment line begin with #
 - valid shell command

Example where F<$HOME/test/subjob1.sh> is a shell script (executable).

 $HOME/test/subjob1.sh
 $HOME/test/subjob2.sh
 $HOME/test/subjob3.sh
 $HOME/test/subjob4.sh
 ...
 $HOME/test/subjob38.sh
 $HOME/test/subjob39.sh
 $HOME/test/subjob40.sh

These jobs could be launch by:

 oarsub -n test -l /core=6,walltime=04:00:00 "oar-parexec -f ./subjob.list.txt"

=head2 Parallel job

You need to put the number of core each small job need with option C<--jobnp>.
If your job is build on OpenMP or MPI,
you can use OAR_NP and OAR_NODE_FILE variables to configure them.
On OAR cluster, you need to use C<oarsh> or a wrapper like C<oar-envsh>
for connexion between node instead of C<ssh>.

Example with parallel small job on 2 core:

 oarsub -n test -l /core=6,walltime=04:00:00 "oar-parexec -j 2 -f ./subjob.list.txt"

=head2 Tracing and master crash

If the master node crash after hours of calculus, everything is lost ?
No, with option C<--logtrace>,
it's possible to remember older result
and not re-run these job the second and next time.

 oarsub -n test -l /core=6,walltime=04:00:00 "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"

After a crash or an C<oardel> command,
you can then re-run the same command that will end to execute the jobs in the list

 oarsub -n test -l /core=6,walltime=04:00:00 "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"

C<logtrace> file are just plain file.
We use the extension '.log' because these files are automatically
eliminate from our backup system!

=head2 Checkpointing and Idempotent

C<oar-parexec> is compatible with the OAR checkpointing.
Il you have 2000 small jobs that need 55h to be done on 6 cores,
you can cut this in small parts.

For this example, we suppose that each small job need about 10min...
So, we send a checkpoint 12min before the end of the process
to let C<oar-parexec> finish the jobs started.
After being checkpointed, C<oar-parexec> do not start any new small job.

 oarsub -t idempotent -n test -l /core=6,walltime=04:00:00 --checkpoint 720 \
   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"

After 3h48min, the OAR job will begin to stop launching new small job.
When all running small job are finished, it's exit.
But as the OAR job is type C<idempotent>,
OAR will re-submit it as long as all small job are not executed...

This way, we let other users a chance to use the cluster!

In this last exemple, we use moldable OAR job with idempotent
to reserve many core for a small time or a few cores for a long time:

 oarsub -t idempotent -n test \
   -l /core=50,walltime=01:05:00 \
   -l /core=6,walltime=04:00:00 \
   --checkpoint 720 \
   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"


=head1 SEE ALSO

oar-dispatch, mpilauncher,
orsh, oar-envsh, ssh


=head1 AUTHORS

Written by Gabriel Moreau, Grenoble - France


=head1 LICENSE AND COPYRIGHT

GPL version 2 or later and Perl equivalent

Copyright (C) 2011 Gabriel Moreau / LEGI - CNRS UMR 5519 - France