X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/036a701b45399e0114eed5b0e1ce6b478c614534..ab5ff91383be492d5a5bcbfdcc008cb32f37a03d:/sdk/cli/bin/crunch-job diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job index 515aaa668d..f092558cd7 100755 --- a/sdk/cli/bin/crunch-job +++ b/sdk/cli/bin/crunch-job @@ -139,7 +139,7 @@ $SIG{'USR2'} = sub my $arv = Arvados->new('apiVersion' => 'v1'); -my $metastream; +my $local_logfile; my $User = $arv->{'users'}->{'current'}->execute; @@ -185,7 +185,7 @@ else $job_id = $Job->{'uuid'}; my $keep_logfile = $job_id . '.log.txt'; -my $local_logfile = File::Temp->new(); +$local_logfile = File::Temp->new(); $Job->{'runtime_constraints'} ||= {}; $Job->{'runtime_constraints'}->{'max_tasks_per_node'} ||= 0; @@ -501,6 +501,27 @@ if (!$have_slurm) # If this job requires a Docker image, install that. my $docker_bin = "/usr/bin/docker.io"; my $docker_image = $Job->{runtime_constraints}->{docker_image} || ""; +if ($docker_image) { + my $docker_pid = fork(); + if ($docker_pid == 0) + { + srun (["srun", "--nodelist=" . join(' ', @node)], + [$docker_bin, 'pull', $docker_image]); + exit ($?); + } + while (1) + { + last if $docker_pid == waitpid (-1, WNOHANG); + freeze_if_want_freeze ($docker_pid); + select (undef, undef, undef, 0.1); + } + # If the Docker image was specified as a hash, pull will fail. + # Ignore that error. We'll see what happens when we try to run later. + if (($? != 0) && ($docker_image !~ /^[0-9a-fA-F]{5,64}$/)) + { + croak("Installing Docker image $docker_image returned exit code $?"); + } +} foreach (qw (script script_version script_parameters runtime_constraints)) { @@ -634,7 +655,7 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++) { $command .= "-e \Q$env_key=$env_val\E "; } - $command .= "$docker_image "; + $command .= "\Q$docker_image\E "; } $command .= "$ENV{CRUNCH_SRC}/crunch_scripts/" . $Job->{"script"}; my @execargs = ('bash', '-c', $command); @@ -925,13 +946,19 @@ sub reapchildren delete $proc{$pid}; # Load new tasks - my $newtask_list = $arv->{'job_tasks'}->{'list'}->execute( - 'where' => { - 'created_by_job_task_uuid' => $Jobstep->{'arvados_task'}->{uuid} - }, - 'order' => 'qsequence' - ); - foreach my $arvados_task (@{$newtask_list->{'items'}}) { + my $newtask_list = []; + my $newtask_results; + do { + $newtask_results = $arv->{'job_tasks'}->{'list'}->execute( + 'where' => { + 'created_by_job_task_uuid' => $Jobstep->{'arvados_task'}->{uuid} + }, + 'order' => 'qsequence', + 'offset' => scalar(@$newtask_list), + ); + push(@$newtask_list, @{$newtask_results->{items}}); + } while (@{$newtask_results->{items}}); + foreach my $arvados_task (@$newtask_list) { my $jobstep = { 'level' => $arvados_task->{'sequence'}, 'failures' => 0, @@ -1224,15 +1251,15 @@ sub Log # ($jobstep_id, $logmessage) $message =~ s{([^ -\176])}{"\\" . sprintf ("%03o", ord($1))}ge; $message .= "\n"; my $datetime; - if ($metastream || -t STDERR) { + if ($local_logfile || -t STDERR) { my @gmtime = gmtime; $datetime = sprintf ("%04d-%02d-%02d_%02d:%02d:%02d", $gmtime[5]+1900, $gmtime[4]+1, @gmtime[3,2,1,0]); } print STDERR ((-t STDERR) ? ($datetime." ".$message) : $message); - if ($metastream) { - print $metastream $datetime . " " . $message; + if ($local_logfile) { + print $local_logfile $datetime . " " . $message; } } @@ -1245,7 +1272,7 @@ sub croak freeze() if @jobstep_todo; collate_output() if @jobstep_todo; cleanup(); - save_meta() if $metastream; + save_meta() if $local_logfile; die; } @@ -1269,6 +1296,7 @@ sub save_meta . quotemeta($local_logfile->filename); my $loglocator = `$cmd`; die "system $cmd failed: $?" if $?; + chomp($loglocator); $local_logfile = undef; # the temp file is automatically deleted Log (undef, "log manifest is $loglocator");