source: trunk/oarutils/oar-parexec @ 82

Last change on this file since 82 was 82, checked in by g7moreau, 12 years ago
  • Delete tab
  • Tmp file more robust name
File size: 17.6 KB
Line 
1#!/usr/bin/perl
2#
3# 2011/11/27 gabriel
4
5use strict;
6
7use Getopt::Long();
8use Pod::Usage;
9use Coro;
10use Coro::Semaphore;
11use Coro::Signal;
12use Coro::Channel;
13use Coro::Handle;
14use IO::File;
15use POSIX qw( WNOHANG WEXITSTATUS );
16use Cwd qw( getcwd );
17
18my $file;
19my $dir;
20my $cmd;
21my $logtrace;
22my $verbose;
23my $job_np         = 1;
24my $nodefile       = $ENV{OAR_NODE_FILE} || '';
25my $masterio;
26my $switchio;
27my $help;
28my $oarsh          = 'oarsh -q -T';
29my $sig_transmit;
30my $sig_checkpoint = 'USR2';
31
32Getopt::Long::GetOptions(
33   'file=s'     => \$file,
34   'dir=s'      => \$dir,
35   'cmd=s'      => \$cmd,
36   'logtrace=s' => \$logtrace,
37   'verbose'    => \$verbose,
38   'help'       => \$help,
39   'oarsh=s'    => \$oarsh,
40   'jobnp=i'    => \$job_np,
41   'nodefile=s' => \$nodefile,
42   'masterio=s' => \$masterio,
43   'switchio'   => \$switchio,
44   'transmit'   => \$sig_transmit,
45   'kill=s'     => \$sig_checkpoint,
46   ) || pod2usage(-verbose => 0);
47pod2usage(-verbose => 2) if $help;
48pod2usage(-verbose => 2) if not (
49 (-e "$file")
50 or (-d "$dir" and $cmd ne '')
51 );
52
53# re-run, keep trace of job already done
54my %state;
55my $log_h = IO::File->new();
56if (-e "$logtrace") {
57   $log_h->open("< $logtrace")
58      or die "error: can't read log file: $!";
59   while (<$log_h>) {
60      $state{$1} = 'start' if m/^start\s+job\s+([^\s]+)\s/;
61      $state{$1} = 'end'   if m/^end\s+job\s+([^\s]+)\s/;
62      }
63   $log_h->close();
64   }
65if ($logtrace) {
66   $log_h->open(">> $logtrace")
67      or die "error: can't append log file $logtrace: $!";
68   $log_h->autoflush;
69   $log_h = unblock $log_h;
70   }
71
72# job to run
73my @job = ();
74if (-e "$file") {
75   my $job_num = 0;
76   open(JOB_LIST, '<', "$file") or die "error: can't open job file $file: $!";
77   while (my $job_cmd = <JOB_LIST>) {
78      chomp $job_cmd;
79      next if $job_cmd =~ m/^#/;
80      next if $job_cmd =~ m/^\s*$/;
81      $job_num++;
82      my ($job_name) = $job_cmd =~ m/#.*?\bname=(\S+?)\b/i;
83      $job_name ||= $job_num;
84      push @job, { name => $job_name, cmd => "$job_cmd" };
85      }
86   close JOB_LIST;
87   }
88else {
89   opendir(DIR, $dir) or die "error: can't open folder $dir: $!";
90   while (my $item = readdir(DIR)) {
91      next if $item =~ m/^\./;
92      next if $item =~ m/:/;
93      next if $item =~ m/\.old$/;
94      next if $item =~ m/\.sav$/;
95      next if $item =~ m/\.bak$/;
96      next if $item =~ m/\.no$/;
97      next unless (-d "$dir/$item");
98      push @job, { name => $item, cmd => "( cd $dir/$item/; $cmd )" };
99      }
100   closedir DIR;
101   }
102
103# ressources available
104my @ressources = ();
105open(NODE_FILE, '<', "$nodefile")
106   or die "can't open $nodefile: $!";
107while (<NODE_FILE>) {
108   chomp;
109   next if m/^#/;
110   next if m/^\s*$/;
111   push @ressources, $_;
112   }
113close NODE_FILE;
114
115my $ressource_size = scalar(@ressources);
116die "error: not enought ressources jobnp $job_np > ressources $ressource_size"
117   if $job_np > $ressource_size;
118
119my $current_dir = getcwd();
120
121my $stderr = $ENV{OAR_STDERR} || '';
122$stderr =~ s/\.stderr$//;
123$stderr = $masterio if $masterio;
124my $stdout = $ENV{OAR_STDOUT} || '';
125$stdout =~ s/\.stdout$//;
126$stdout = $masterio if $masterio;
127
128my $finished = new Coro::Signal;
129my $job_todo = new Coro::Semaphore 0;
130my $job_name_maxlen;
131for (@job) {
132   $job_todo->up;
133   $job_name_maxlen = length($_->{name}) if length($_->{name}) > $job_name_maxlen;
134   }
135
136# slice of ressources for parallel job
137my $ressources = new Coro::Channel;
138for my $slot (1 .. int($ressource_size / $job_np)) {
139   $ressources->put(
140      join(',',
141         @ressources[ (($slot - 1) * $job_np) .. (($slot * $job_np) - 1) ])
142         );
143   }
144
145my %scheduled = ();
146
147# OAR checkpoint and default signal SIGUSR2
148my $oar_checkpoint = new Coro::Semaphore 0;
149$SIG{$sig_checkpoint} = sub {
150   print "warning: receive checkpoint at "
151      . time
152      . ", no new job, just finishing running job\n"
153      if $verbose;
154   $oar_checkpoint->up();
155   };
156
157# asynchrone notify job
158async {
159   while () {
160      next if $oar_checkpoint->count() == 0;
161
162      # only notify with transmit flag
163      if (not $sig_transmit) {
164         cede;
165         next;
166         }
167
168      for my $job_pid (keys %scheduled) {
169         my $job_name     = $scheduled{$job_pid}->{name};
170         my $job_pidfile  = $scheduled{$job_pid}->{pidfile};
171         my $node_connect = $scheduled{$job_pid}->{node_connect};
172
173         my $fh = IO::File->new();
174         $fh->open("| $oarsh $node_connect >/dev/null 2>&1")
175            or die "error: can't notify subjob: $!";
176
177         $fh->autoflush;
178         $fh = unblock $fh;
179
180         $fh->print("kill -$sig_checkpoint \$(cat $job_pidfile)");
181         $fh->print("exit\n");
182
183         print "warning: transmit signal $sig_checkpoint"
184            . " to $job_name on $node_connect.\n"
185            if $verbose;
186
187         close $fh;
188         cede;
189         }
190      }
191   }
192
193# asynchrone start job block
194async {
195   JOB:
196   for my $job (@job) {
197      my $job_name = $job->{name};
198      my $job_cmd  = $job->{cmd};
199
200      # job has been already run ?
201      if (exists $state{$job_name}) {
202         if ($state{$job_name} eq 'start') {
203            print "warning: job $job_name was not clearly finished, relaunching...\n"
204               if $verbose;
205            }
206         elsif ($state{$job_name} eq 'end') {
207            delete $state{$job_name}; # free memory
208            $job_todo->down;
209            print "warning: job $job_name already run\n" if $verbose;
210            cede;
211            next JOB;
212            }
213         }
214
215      # take job ressource
216      my $job_ressource = $ressources->get;
217
218      # no more launch job when OAR checkpointing
219      last JOB if $oar_checkpoint->count() > 0;
220
221      my ($node_connect) = split ',', $job_ressource;
222      my $fh = IO::File->new();
223      my $job_pid = $fh->open("| $oarsh $node_connect >/dev/null 2>&1")
224         or die "error: can't start subjob: $!";
225
226      $fh->autoflush;
227      $fh = unblock $fh;
228
229      my $msg = sprintf "start job %${job_name_maxlen}s / %5i at %s on node %s\n",
230         $job_name, $job_pid, time, $job_ressource;
231      $log_h->print($msg) if $logtrace;
232      print($msg) if $verbose;
233
234      my ($job_stdout, $job_stderr);
235      $job_stdout = ">  $stdout-$job_name.stdout" if $stdout ne '' and $switchio;
236      $job_stderr = "2> $stderr-$job_name.stderr" if $stderr ne '' and $switchio;
237
238      my $job_nodefile = "/tmp/oar-parexec-$ENV{LOGNAME}-$ENV{OAR_JOB_ID}-$job_name";
239      my $job_pidfile  = "/tmp/oar-parexec-$ENV{LOGNAME}-$ENV{OAR_JOB_ID}-$job_name.pid";
240
241      $scheduled{$job_pid} = {
242         fh           => $fh,
243         node_connect => $node_connect,
244         ressource    => $job_ressource,
245         name         => $job_name,
246         pidfile      => $job_pidfile,
247         };
248
249      # set job environment, run it and clean
250      if ($job_np > 1) {
251         $fh->print("printf \""
252               . join('\n', split(',', $job_ressource,))
253               . "\" > $job_nodefile\n");
254         $fh->print("OAR_NODE_FILE=$job_nodefile\n");
255         $fh->print("OAR_NP=$job_np\n");
256         $fh->print("export OAR_NODE_FILE\n");
257         $fh->print("export OAR_NP\n");
258         $fh->print("unset OAR_MSG_NODEFILE\n");
259         }
260      $fh->print("cd $current_dir\n");
261      if ($sig_transmit) {
262         $fh->print("trap 'kill -$sig_checkpoint \$(jobs -p)' $sig_checkpoint\n");
263         $fh->print("echo \$\$ > $job_pidfile\n");
264         $fh->print("$job_cmd $job_stdout $job_stderr &\n");
265         $fh->print("while [ \$(jobs -p | wc -l) -gt 0 ]\n");
266         $fh->print("do\n");
267         $fh->print("   wait\n");
268         $fh->print("done\n");
269         $fh->print("rm -f $job_pidfile\n");
270         }
271      else {
272         $fh->print("$job_cmd $job_stdout $job_stderr\n");
273         }
274      $fh->print("rm -f $job_nodefile\n") if $job_np > 1;
275      $fh->print("exit\n");
276      cede;
277      }
278   }
279
280# asynchrone end job block
281async {
282   while () {
283      for my $job_pid (keys %scheduled) {
284         # non blocking PID test
285         if (waitpid($job_pid, WNOHANG)) {
286            my $msg = sprintf "end   job %${job_name_maxlen}s / %5i at %s on node %s\n",
287               $scheduled{$job_pid}->{name},
288               $job_pid, time, $scheduled{$job_pid}->{ressource};
289
290            # Job non finish, just suspend if received checkpoint signal
291            $msg =~ s/^end\s+job/suspend job/
292               if $sig_transmit and $oar_checkpoint->count() > 0;
293
294            $log_h->print($msg) if $logtrace;
295            print($msg) if $verbose;
296            close $scheduled{$job_pid}->{fh};
297            # leave ressources for another job
298            $ressources->put($scheduled{$job_pid}->{ressource});
299            $job_todo->down;
300            delete $scheduled{$job_pid};
301            }
302         cede;
303         }
304
305      # checkpointing ! just finishing running job and quit
306      $finished->send if $oar_checkpoint->count() > 0 and scalar(keys(%scheduled)) == 0;
307
308      $finished->send if $job_todo->count() == 0;
309      cede;
310      }
311   }
312
313cede;
314
315# all job have been done
316$finished->wait;
317
318# close log trace file
319$log_h->close() if $logtrace;
320
321__END__
322
323=head1 NAME
324
325oar-parexec - parallel execution of many small job
326
327=head1 SYNOPSIS
328
329 oar-parexec --file filecommand \
330    [--logtrace tracefile] [--verbose] \
331    [--jobnp integer] [--nodefile filenode] [--oarsh sssh] \
332    [--switchio] [--masterio basefileio]
333
334 oar-parexec --dir foldertoiterate --cmd commandtolaunch \
335    [--logtrace tracefile] [--verbose] \
336    [--jobnp integer] [--nodefile filenode] [--oarsh sssh] \
337    [--switchio] [--masterio basefileio]
338
339 oar-parexec --help
340
341=head1 DESCRIPTION
342
343C<oar-parexec> can execute lot of small job in parallel inside a cluster.
344Number of parallel job at one time cannot exceed the number of core define in the node file
345C<oar-parexec> is easier to use inside an OAR job environment
346which define automatically these strategics parameters...
347However, it can be used outside OAR.
348
349Option C<--file> or C<--dir> and C<--cmd> are the only mandatory parameters.
350
351Small job will be launch in the same folder as the master job.
352Two environment variable are defined for each small job
353and only in case of parallel small job (option C<--jobnp> > 1).
354
355 OAR_NODE_FILE - file that list node for parallel computing
356 OAR_NP        - number of processor affected
357
358The file define by OAR_NODE_FILE is created  in /tmp
359on the node before launching the small job
360and this file will be delete after job complete.
361C<oar-parexec> is a simple script,
362OAR_NODE_FILE will not be deleted in case of crash of the master job.
363
364OAR define other variable that are equivalent to OAR_NODE_FILE:
365OAR_NODEFILE, OAR_FILE_NODES, OAR_RESOURCE_FILE...
366You can use in your script the OAR original file ressources
367by using these variable if you need it.
368
369
370=head1 OPTIONS
371
372=over 12
373
374=item B<-f|--file filecommand>
375
376File name which content job list.
377For the JOB_NAME definition,
378the first valid job in the list will have the number 1 and so on...
379
380It's possible to fix the name inside a comment on the job line.
381For example:
382
383 $HOME/test/subjob1.sh # name=subjob1
384
385The key C<name> is case insensitive,
386the associated value cannot have a space...
387
388=item B<-d|--dir foldertoiterate>
389
390Command C<--cmd> will be launch in all sub-folder of this master folder.
391Files in this folder will be ignored.
392Sub-folder name which begin with F<.>
393or finish with F<.old>, F<.sav>, F<.bak>, F<.no> will either be ignored...
394
395The JOB_NAME is simply the Sub-folder name.
396
397=item B<-c|--cmd commandtolaunch>
398
399Command (and argument to it) tha will be launch in all sub-folder
400parameter folfer C<--dir>
401
402=item B<-l|--logtrace tracefile>
403
404File which log and trace running job.
405In case of running the same master command (after crash for example),
406only job that are not mark as done will be run again.
407Be careful, job mark as running (start but not finish) will be run again.
408Tracing is base on the JOB_NAME between multiple run.
409
410This option is very usefull in case of crash
411but also for checkpointing and idempotent OAR job.
412
413=item B<-v|--verbose>
414
415=item B<-j|--jobnp integer>
416
417Number of processor to allocated for each small job.
4181 by default.
419
420=item B<-n|--nodefile filenode>
421
422File name that list all the node where job could be launch.
423By defaut, it's define automatically by OAR via
424environment variable C<OAR_NODE_FILE>.
425
426For example, if you want to use 6 core on your cluster node,
427you need to put 6 times the hostname node in this file,
428one per line...
429It's a very common file in MPI process !
430
431=item B<-o|-oarsh command>
432
433Command use to launch a shell on a node.
434By default
435
436 oarsh -q -T
437
438Change it to C<ssh> if you are not using an OAR cluster...
439
440=item B<-s|--switchio>
441
442Each small job will have it's own output STDOUT and STDERR
443base on master OAR job with C<JOB_NAME> inside
444(or base on C<basefileio> if option C<masterio>).
445Example :
446
447 OAR.151524.stdout -> OAR.151524-JOB_NAME.stdout
448
449where 151524 here is the master C<OAR_JOB_ID>
450and C<JOB_NAME> is the small job name.
451
452=item B<-m|--masterio basefileio>
453
454The C<basefileio> will be use in place of environment variable
455C<OAR_STDOUT> and C<OAR_STDERR> (without extension) to build the base name of the small job standart output
456(only use when option C<swithio> is activated).
457
458=item B<-k|--kill signal>
459
460Signal to listen and make a clean stop of the current C<oar-parexec> process.
461By default, use USR2 signal (see C<kill -l>> for a list of possible signal).
462
463=item B<-t|--transmit>
464
465Resend catch signal to sub-job when receiving it.
466By default, no signal is transmis to child process.
467
468It's only valuable if use for long sub-job than can
469in return make themselves a clean restart.
470
471
472=item B<-h|--help>
473
474=back
475
476
477=head1 EXAMPLE
478
479=head2 Simple list of sequential job
480
481Content for the job file command (option C<--file>) could have:
482
483 - empty line
484 - comment line begin with #
485 - valid shell command
486
487Example where F<$HOME/test/subjob1.sh> is a shell script (executable).
488
489 $HOME/test/subjob01.sh  # name=subjob01
490 $HOME/test/subjob02.sh  # name=subjob02
491 $HOME/test/subjob03.sh  # name=subjob03
492 $HOME/test/subjob04.sh  # name=subjob04
493 ...
494 $HOME/test/subjob38.sh  # name=subjob38
495 $HOME/test/subjob39.sh  # name=subjob39
496 $HOME/test/subjob40.sh  # name=subjob40
497
498These jobs could be launch by:
499
500 oarsub -n test -l /core=6,walltime=04:00:00 \
501   "oar-parexec -f ./subjob.list.txt"
502
503=head2 Folder job
504
505In a folder F<subjob.d>, create sub-folder with your data inside : F<test1>, <test2>...
506The same command will be executed in every sub-folder.
507C<oar-parexec> change the current directory to the sub-folder before launching it.
508
509A very simple job could be:
510
511 oarsub -n test -l /core=6,walltime=04:00:00 \
512   "oar-parexec -d ./subjob.d -c 'sleep 10; env'"
513
514The command C<env> will be excuted in all folder F<test1>, F<test2>... after a 10s pause.
515
516Sometime, it's simpler to use file list command,
517sometime, jobs by folder with the same command run is more relevant.
518
519=head2 Parallel job
520
521You need to put the number of core each small job need with option C<--jobnp>.
522If your job is build on OpenMP or MPI,
523you can use OAR_NP and OAR_NODE_FILE variables to configure them.
524On OAR cluster, you need to use C<oarsh> or a wrapper like C<oar-envsh>
525for connexion between node instead of C<ssh>.
526
527Example with parallel small job on 2 core:
528
529 oarsub -n test -l /core=6,walltime=04:00:00 \
530   "oar-parexec -j 2 -f ./subjob.list.txt"
531
532=head2 Tracing and master crash
533
534If the master node crash after hours of calculus, everything is lost ?
535No, with option C<--logtrace>,
536it's possible to remember older result
537and not re-run these job the second and next time.
538
539 oarsub -n test -l /core=6,walltime=04:00:00 \
540   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"
541
542After a crash or an C<oardel> command,
543you can then re-run the same command that will end to execute the jobs in the list
544
545 oarsub -n test -l /core=6,walltime=04:00:00 \
546   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"
547
548C<logtrace> file are just plain file.
549We use the extension '.log' because these files are automatically
550eliminate from our backup system!
551
552=head2 Checkpointing and Idempotent
553
554C<oar-parexec> is compatible with the OAR checkpointing.
555Il you have 2000 small jobs that need 55h to be done on 6 cores,
556you can cut this in small parts.
557
558For this example, we suppose that each small job need about 10min...
559So, we send a checkpoint 12min before the end of the process
560to let C<oar-parexec> finish the jobs started.
561After being checkpointed, C<oar-parexec> do not start any new small job.
562
563 oarsub -t idempotent -n test \
564   -l /core=6,walltime=04:00:00 \
565   --checkpoint 720 \
566   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"
567
568After 3h48min, the OAR job will begin to stop launching new small job.
569When all running small job are finished, it's exit.
570But as the OAR job is type C<idempotent>,
571OAR will re-submit it as long as all small job are not executed...
572
573This way, we let other users a chance to use the cluster!
574
575In this last exemple, we use moldable OAR job with idempotent
576to reserve many core for a small time or a few cores for a long time:
577
578 oarsub -t idempotent -n test \
579   -l /core=50,walltime=01:05:00 \
580   -l /core=6,walltime=04:00:00 \
581   --checkpoint 720 \
582   "oar-parexec -f ./subjob.list.txt -l ./subjob.list.log"
583
584=head2 Signal, recurse and long job
585
586By default, OAR use signal USR2 for checkpointing.
587It's possible to change this with option C<--kill>.
588
589When use with long small job, checkpointing could be too long...
590More than walltime!
591The option C<--transmit> could be use to checkpoint small job!
592These long small job will then stop cleanly and will be restarted next time.
593
594In the C<logtrace> file, small job will have the status suspend.
595They will be launch with the same command line at the next OAR run.
596
597=head1 SEE ALSO
598
599oar-dispatch, mpilauncher,
600orsh, oar-envsh, ssh
601
602
603=head1 AUTHORS
604
605Written by Gabriel Moreau, Grenoble - France
606
607
608=head1 LICENSE AND COPYRIGHT
609
610GPL version 2 or later and Perl equivalent
611
612Copyright (C) 2011 Gabriel Moreau / LEGI - CNRS UMR 5519 - France
613
Note: See TracBrowser for help on using the repository browser.