2 # -*- mode: perl; perl-indent-level: 2; indent-tabs-mode: nil; -*-
6 crunch-job: Execute job steps, save snapshots as requested, collate output.
10 Obtain job details from Arvados, run tasks on compute nodes (typically
11 invoked by scheduler on controller):
13 crunch-job --job x-y-z --git-dir /path/to/repo/.git
15 Obtain job details from command line, run tasks on local machine
16 (typically invoked by application or developer on VM):
18 crunch-job --job '{"script_version":"/path/to/working/tree","script":"scriptname",...}'
20 crunch-job --job '{"repository":"https://github.com/curoverse/arvados.git","script_version":"master","script":"scriptname",...}'
28 If the job is already locked, steal the lock and run it anyway.
32 Path to a .git directory (or a git URL) where the commit given in the
33 job's C<script_version> attribute is to be found. If this is I<not>
34 given, the job's C<repository> attribute will be used.
38 Arvados API authorization token to use during the course of the job.
42 Do not clear per-job/task temporary directories during initial job
43 setup. This can speed up development and debugging when running jobs
48 UUID of the job to run, or a JSON-encoded job resource without a
49 UUID. If the latter is given, a new job object will be created.
53 =head1 RUNNING JOBS LOCALLY
55 crunch-job's log messages appear on stderr along with the job tasks'
56 stderr streams. The log is saved in Keep at each checkpoint and when
59 If the job succeeds, the job's output locator is printed on stdout.
61 While the job is running, the following signals are accepted:
65 =item control-C, SIGINT, SIGQUIT
67 Save a checkpoint, terminate any job tasks that are running, and stop.
71 Save a checkpoint and continue.
75 Refresh node allocation (i.e., check whether any nodes have been added
76 or unallocated) and attributes of the Job record that should affect
77 behavior (e.g., cancel job if cancelled_at becomes non-nil).
85 use POSIX ':sys_wait_h';
86 use POSIX qw(strftime);
87 use Fcntl qw(F_GETFL F_SETFL O_NONBLOCK);
91 use Digest::MD5 qw(md5_hex);
97 use File::Path qw( make_path remove_tree );
99 use constant TASK_TEMPFAIL => 111;
100 use constant EX_TEMPFAIL => 75;
101 use constant EX_RETRY_UNLOCKED => 93;
103 $ENV{"TMPDIR"} ||= "/tmp";
104 unless (defined $ENV{"CRUNCH_TMP"}) {
105 $ENV{"CRUNCH_TMP"} = $ENV{"TMPDIR"} . "/crunch-job";
106 if ($ENV{"USER"} ne "crunch" && $< != 0) {
107 # use a tmp dir unique for my uid
108 $ENV{"CRUNCH_TMP"} .= "-$<";
112 # Create the tmp directory if it does not exist
113 if ( ! -d $ENV{"CRUNCH_TMP"} ) {
114 make_path $ENV{"CRUNCH_TMP"} or die "Failed to create temporary working directory: " . $ENV{"CRUNCH_TMP"};
117 $ENV{"JOB_WORK"} = $ENV{"CRUNCH_TMP"} . "/work";
118 $ENV{"CRUNCH_INSTALL"} = "$ENV{CRUNCH_TMP}/opt";
119 $ENV{"CRUNCH_WORK"} = $ENV{"JOB_WORK"}; # deprecated
120 mkdir ($ENV{"JOB_WORK"});
129 my $docker_bin = "docker.io";
130 GetOptions('force-unlock' => \$force_unlock,
131 'git-dir=s' => \$git_dir,
132 'job=s' => \$jobspec,
133 'job-api-token=s' => \$job_api_token,
134 'no-clear-tmp' => \$no_clear_tmp,
135 'resume-stash=s' => \$resume_stash,
136 'docker-bin=s' => \$docker_bin,
139 if (defined $job_api_token) {
140 $ENV{ARVADOS_API_TOKEN} = $job_api_token;
143 my $have_slurm = exists $ENV{SLURM_JOB_ID} && exists $ENV{SLURM_NODELIST};
148 $main::ENV{CRUNCH_DEBUG} = 1;
152 $main::ENV{CRUNCH_DEBUG} = 0;
155 my $arv = Arvados->new('apiVersion' => 'v1');
164 if ($jobspec =~ /^[-a-z\d]+$/)
166 # $jobspec is an Arvados UUID, not a JSON job specification
167 $Job = api_call("jobs/get", uuid => $jobspec);
172 $local_job = JSON::decode_json($jobspec);
176 # Make sure our workers (our slurm nodes, localhost, or whatever) are
177 # at least able to run basic commands: they aren't down or severely
180 if (($Job || $local_job)->{docker_image_locator}) {
181 $cmd = [$docker_bin, 'ps', '-q'];
183 Log(undef, "Sanity check is `@$cmd`");
184 srun(["srun", "--nodes=\Q$ENV{SLURM_NNODES}\E", "--ntasks-per-node=1"],
188 Log(undef, "Sanity check failed: ".exit_status_s($?));
191 Log(undef, "Sanity check OK");
194 my $User = api_call("users/current");
197 if (!$force_unlock) {
198 # Claim this job, and make sure nobody else does
199 eval { api_call("jobs/lock", uuid => $Job->{uuid}); };
201 Log(undef, "Error while locking job, exiting ".EX_TEMPFAIL);
210 map { croak ("No $_ specified") unless $local_job->{$_} }
211 qw(script script_version script_parameters);
214 $local_job->{'is_locked_by_uuid'} = $User->{'uuid'};
215 $local_job->{'started_at'} = gmtime;
216 $local_job->{'state'} = 'Running';
218 $Job = api_call("jobs/create", job => $local_job);
220 $job_id = $Job->{'uuid'};
222 my $keep_logfile = $job_id . '.log.txt';
223 log_writer_start($keep_logfile);
225 $Job->{'runtime_constraints'} ||= {};
226 $Job->{'runtime_constraints'}->{'max_tasks_per_node'} ||= 0;
227 my $max_ncpus = $Job->{'runtime_constraints'}->{'max_tasks_per_node'};
229 my $gem_versions = `gem list --quiet arvados-cli 2>/dev/null`;
231 $gem_versions =~ s/^arvados-cli \(/ with arvados-cli Gem version(s) /;
232 chomp($gem_versions);
233 chop($gem_versions); # Closing parentheses
238 "running from " . ((-e $0) ? realpath($0) : "stdin") . $gem_versions);
240 Log (undef, "check slurm allocation");
243 # Should use $ENV{SLURM_TASKS_PER_NODE} instead of sinfo? (eg. "4(x3),2,4(x2)")
247 my $localcpus = 0 + `grep -cw ^processor /proc/cpuinfo` || 1;
248 push @sinfo, "$localcpus localhost";
250 if (exists $ENV{SLURM_NODELIST})
252 push @sinfo, `sinfo -h --format='%c %N' --nodes=\Q$ENV{SLURM_NODELIST}\E`;
256 my ($ncpus, $slurm_nodelist) = split;
257 $ncpus = $max_ncpus if $max_ncpus && $ncpus > $max_ncpus;
260 while ($slurm_nodelist =~ s/^([^\[,]+?(\[.*?\])?)(,|$)//)
263 if ($nodelist =~ /\[((\d+)(-(\d+))?(,(\d+)(-(\d+))?)*)\]/)
266 foreach (split (",", $ranges))
279 push @nodelist, map {
281 $n =~ s/\[[-,\d]+\]/$_/;
288 push @nodelist, $nodelist;
291 foreach my $nodename (@nodelist)
293 Log (undef, "node $nodename - $ncpus slots");
294 my $node = { name => $nodename,
296 # The number of consecutive times a task has been dispatched
297 # to this node and failed.
299 # The number of consecutive times that SLURM has reported
300 # a node failure since the last successful task.
302 # Don't dispatch work to this node until this time
303 # (in seconds since the epoch) has passed.
305 foreach my $cpu (1..$ncpus)
307 push @slot, { node => $node,
311 push @node, @nodelist;
316 # Ensure that we get one jobstep running on each allocated node before
317 # we start overloading nodes with concurrent steps
319 @slot = sort { $a->{cpu} <=> $b->{cpu} } @slot;
322 $Job->update_attributes(
323 'tasks_summary' => { 'failed' => 0,
328 Log (undef, "start");
329 $SIG{'INT'} = sub { $main::please_freeze = 1; };
330 $SIG{'QUIT'} = sub { $main::please_freeze = 1; };
331 $SIG{'TERM'} = \&croak;
332 $SIG{'TSTP'} = sub { $main::please_freeze = 1; };
333 $SIG{'ALRM'} = sub { $main::please_info = 1; };
334 $SIG{'CONT'} = sub { $main::please_continue = 1; };
335 $SIG{'HUP'} = sub { $main::please_refresh = 1; };
337 $main::please_freeze = 0;
338 $main::please_info = 0;
339 $main::please_continue = 0;
340 $main::please_refresh = 0;
341 my $jobsteps_must_output_keys = 0; # becomes 1 when any task outputs a key
343 grep { $ENV{$1} = $2 if /^(NOCACHE.*?)=(.*)/ } split ("\n", $$Job{knobs});
344 $ENV{"CRUNCH_JOB_UUID"} = $job_id;
345 $ENV{"JOB_UUID"} = $job_id;
348 my @jobstep_todo = ();
349 my @jobstep_done = ();
350 my @jobstep_tomerge = ();
351 my $jobstep_tomerge_level = 0;
352 my $squeue_checked = 0;
353 my $latest_refresh = scalar time;
357 if (defined $Job->{thawedfromkey})
359 thaw ($Job->{thawedfromkey});
363 my $first_task = api_call("job_tasks/create", job_task => {
364 'job_uuid' => $Job->{'uuid'},
369 push @jobstep, { 'level' => 0,
371 'arvados_task' => $first_task,
373 push @jobstep_todo, 0;
379 must_lock_now("$ENV{CRUNCH_TMP}/.lock", "a job is already running here.");
382 my $build_script = handle_readall(\*DATA);
383 my $nodelist = join(",", @node);
384 my $git_tar_count = 0;
386 if (!defined $no_clear_tmp) {
387 # Clean out crunch_tmp/work, crunch_tmp/opt, crunch_tmp/src*
388 Log (undef, "Clean work dirs");
390 my $cleanpid = fork();
393 # Find FUSE mounts under $CRUNCH_TMP and unmount them.
394 # Then clean up work directories.
395 # TODO: When #5036 is done and widely deployed, we can limit mount's
396 # -t option to simply fuse.keep.
397 srun (["srun", "--nodelist=$nodelist", "-D", $ENV{'TMPDIR'}],
398 ['bash', '-ec', '-o', 'pipefail', 'mount -t fuse,fuse.keep | awk "(index(\$3, \"$CRUNCH_TMP\") == 1){print \$3}" | xargs -r -n 1 fusermount -u -z; sleep 1; rm -rf $JOB_WORK $CRUNCH_INSTALL $CRUNCH_TMP/task $CRUNCH_TMP/src* $CRUNCH_TMP/*.cid']);
403 last if $cleanpid == waitpid (-1, WNOHANG);
404 freeze_if_want_freeze ($cleanpid);
405 select (undef, undef, undef, 0.1);
408 Log(undef, "Clean work dirs: exit ".exit_status_s($?));
409 exit(EX_RETRY_UNLOCKED);
413 # If this job requires a Docker image, install that.
414 my ($docker_locator, $docker_stream, $docker_hash, $docker_limitmem, $dockeruserarg);
415 if ($docker_locator = $Job->{docker_image_locator}) {
416 ($docker_stream, $docker_hash) = find_docker_image($docker_locator);
419 croak("No Docker image hash found from locator $docker_locator");
421 $docker_stream =~ s/^\.//;
422 my $docker_install_script = qq{
423 if ! $docker_bin images -q --no-trunc --all | grep -qxF \Q$docker_hash\E; then
424 arv-get \Q$docker_locator$docker_stream/$docker_hash.tar\E | $docker_bin load
427 my $docker_pid = fork();
428 if ($docker_pid == 0)
430 srun (["srun", "--nodelist=" . join(',', @node)],
431 ["/bin/sh", "-ec", $docker_install_script]);
436 last if $docker_pid == waitpid (-1, WNOHANG);
437 freeze_if_want_freeze ($docker_pid);
438 select (undef, undef, undef, 0.1);
442 croak("Installing Docker image from $docker_locator exited "
446 # Determine whether this version of Docker supports memory+swap limits.
447 srun(["srun", "--nodelist=" . $node[0]],
448 ["/bin/sh", "-ec", "$docker_bin run --help | grep -qe --memory-swap="],
450 $docker_limitmem = ($? == 0);
452 # Find a non-root Docker user to use.
453 # Tries the default user for the container, then 'crunch', then 'nobody',
454 # testing for whether the actual user id is non-zero. This defends against
455 # mistakes but not malice, but we intend to harden the security in the future
456 # so we don't want anyone getting used to their jobs running as root in their
458 my @tryusers = ("", "crunch", "nobody");
459 foreach my $try_user (@tryusers) {
461 if ($try_user eq "") {
462 Log(undef, "Checking if container default user is not UID 0");
465 Log(undef, "Checking if user '$try_user' is not UID 0");
466 $try_user_arg = "--user=$try_user";
468 srun(["srun", "--nodelist=" . $node[0]],
470 "a=`$docker_bin run --rm $try_user_arg $docker_hash id --user` && " .
474 $dockeruserarg = $try_user_arg;
475 if ($try_user eq "") {
476 Log(undef, "Container will run with default user");
478 Log(undef, "Container will run with $dockeruserarg");
484 if (!defined $dockeruserarg) {
485 croak("Could not find a user in container that is not UID 0 (tried default user, @tryusers) or there was a problem running 'id' in the container.");
488 if ($Job->{arvados_sdk_version}) {
489 # The job also specifies an Arvados SDK version. Add the SDKs to the
490 # tar file for the build script to install.
491 Log(undef, sprintf("Packing Arvados SDK version %s for installation",
492 $Job->{arvados_sdk_version}));
493 add_git_archive("git", "--git-dir=$git_dir", "archive",
494 "--prefix=.arvados.sdk/",
495 $Job->{arvados_sdk_version}, "sdk");
499 if (!defined $git_dir && $Job->{'script_version'} =~ m{^/}) {
500 # If script_version looks like an absolute path, *and* the --git-dir
501 # argument was not given -- which implies we were not invoked by
502 # crunch-dispatch -- we will use the given path as a working
503 # directory instead of resolving script_version to a git commit (or
504 # doing anything else with git).
505 $ENV{"CRUNCH_SRC_COMMIT"} = $Job->{'script_version'};
506 $ENV{"CRUNCH_SRC"} = $Job->{'script_version'};
509 # Resolve the given script_version to a git commit sha1. Also, if
510 # the repository is remote, clone it into our local filesystem: this
511 # ensures "git archive" will work, and is necessary to reliably
512 # resolve a symbolic script_version like "master^".
513 $ENV{"CRUNCH_SRC"} = "$ENV{CRUNCH_TMP}/src";
515 Log (undef, "Looking for version ".$Job->{script_version}." from repository ".$Job->{repository});
517 $ENV{"CRUNCH_SRC_COMMIT"} = $Job->{script_version};
519 # If we're running under crunch-dispatch, it will have already
520 # pulled the appropriate source tree into its own repository, and
521 # given us that repo's path as $git_dir.
523 # If we're running a "local" job, we might have to fetch content
524 # from a remote repository.
526 # (Currently crunch-dispatch gives a local path with --git-dir, but
527 # we might as well accept URLs there too in case it changes its
529 my $repo = $git_dir || $Job->{'repository'};
531 # Repository can be remote or local. If remote, we'll need to fetch it
532 # to a local dir before doing `git log` et al.
535 if ($repo =~ m{://|^[^/]*:}) {
536 # $repo is a git url we can clone, like git:// or https:// or
537 # file:/// or [user@]host:repo.git. Note "user/name@host:foo" is
538 # not recognized here because distinguishing that from a local
539 # path is too fragile. If you really need something strange here,
540 # use the ssh:// form.
541 $repo_location = 'remote';
542 } elsif ($repo =~ m{^\.*/}) {
543 # $repo is a local path to a git index. We'll also resolve ../foo
544 # to ../foo/.git if the latter is a directory. To help
545 # disambiguate local paths from named hosted repositories, this
546 # form must be given as ./ or ../ if it's a relative path.
547 if (-d "$repo/.git") {
548 $repo = "$repo/.git";
550 $repo_location = 'local';
552 # $repo is none of the above. It must be the name of a hosted
554 my $arv_repo_list = api_call("repositories/list",
555 'filters' => [['name','=',$repo]]);
556 my @repos_found = @{$arv_repo_list->{'items'}};
557 my $n_found = $arv_repo_list->{'serverResponse'}->{'items_available'};
559 Log(undef, "Repository '$repo' -> "
560 . join(", ", map { $_->{'uuid'} } @repos_found));
563 croak("Error: Found $n_found repositories with name '$repo'.");
565 $repo = $repos_found[0]->{'fetch_url'};
566 $repo_location = 'remote';
568 Log(undef, "Using $repo_location repository '$repo'");
569 $ENV{"CRUNCH_SRC_URL"} = $repo;
571 # Resolve given script_version (we'll call that $treeish here) to a
572 # commit sha1 ($commit).
573 my $treeish = $Job->{'script_version'};
575 if ($repo_location eq 'remote') {
576 # We minimize excess object-fetching by re-using the same bare
577 # repository in CRUNCH_TMP/.git for multiple crunch-jobs -- we
578 # just keep adding remotes to it as needed.
579 my $local_repo = $ENV{'CRUNCH_TMP'}."/.git";
580 my $gitcmd = "git --git-dir=\Q$local_repo\E";
582 # Set up our local repo for caching remote objects, making
584 if (!-d $local_repo) {
585 make_path($local_repo) or croak("Error: could not create $local_repo");
587 # This works (exits 0 and doesn't delete fetched objects) even
588 # if $local_repo is already initialized:
589 `$gitcmd init --bare`;
591 croak("Error: $gitcmd init --bare exited ".exit_status_s($?));
594 # If $treeish looks like a hash (or abbrev hash) we look it up in
595 # our local cache first, since that's cheaper. (We don't want to
596 # do that with tags/branches though -- those change over time, so
597 # they should always be resolved by the remote repo.)
598 if ($treeish =~ /^[0-9a-f]{7,40}$/s) {
599 # Hide stderr because it's normal for this to fail:
600 my $sha1 = `$gitcmd rev-list -n1 ''\Q$treeish\E 2>/dev/null`;
602 # Careful not to resolve a branch named abcdeff to commit 1234567:
603 $sha1 =~ /^$treeish/ &&
604 $sha1 =~ /^([0-9a-f]{40})$/s) {
606 Log(undef, "Commit $commit already present in $local_repo");
610 if (!defined $commit) {
611 # If $treeish isn't just a hash or abbrev hash, or isn't here
612 # yet, we need to fetch the remote to resolve it correctly.
614 # First, remove all local heads. This prevents a name that does
615 # not exist on the remote from resolving to (or colliding with)
616 # a previously fetched branch or tag (possibly from a different
618 remove_tree("$local_repo/refs/heads", {keep_root => 1});
620 Log(undef, "Fetching objects from $repo to $local_repo");
621 `$gitcmd fetch --no-progress --tags ''\Q$repo\E \Q+refs/heads/*:refs/heads/*\E`;
623 croak("Error: `$gitcmd fetch` exited ".exit_status_s($?));
627 # Now that the data is all here, we will use our local repo for
628 # the rest of our git activities.
632 my $gitcmd = "git --git-dir=\Q$repo\E";
633 my $sha1 = `$gitcmd rev-list -n1 ''\Q$treeish\E`;
634 unless ($? == 0 && $sha1 =~ /^([0-9a-f]{40})$/) {
635 croak("`$gitcmd rev-list` exited "
637 .", '$treeish' not found, giving up");
640 Log(undef, "Version $treeish is commit $commit");
642 if ($commit ne $Job->{'script_version'}) {
643 # Record the real commit id in the database, frozentokey, logs,
644 # etc. -- instead of an abbreviation or a branch name which can
645 # become ambiguous or point to a different commit in the future.
646 if (!$Job->update_attributes('script_version' => $commit)) {
647 croak("Error: failed to update job's script_version attribute");
651 $ENV{"CRUNCH_SRC_COMMIT"} = $commit;
652 add_git_archive("$gitcmd archive ''\Q$commit\E");
655 my $git_archive = combined_git_archive();
656 if (!defined $git_archive) {
657 Log(undef, "Skip install phase (no git archive)");
659 Log(undef, "Warning: This probably means workers have no source tree!");
664 my $install_script_tries_left = 3;
665 for (my $attempts = 0; $attempts < 3; $attempts++) {
666 Log(undef, "Run install script on all workers");
668 my @srunargs = ("srun",
669 "--nodelist=$nodelist",
670 "-D", $ENV{'TMPDIR'}, "--job-name=$job_id");
671 my @execargs = ("sh", "-c",
672 "mkdir -p $ENV{CRUNCH_INSTALL} && cd $ENV{CRUNCH_TMP} && perl -");
674 $ENV{"CRUNCH_GIT_ARCHIVE_HASH"} = md5_hex($git_archive);
675 my ($install_stderr_r, $install_stderr_w);
676 pipe $install_stderr_r, $install_stderr_w or croak("pipe() failed: $!");
677 set_nonblocking($install_stderr_r);
678 my $installpid = fork();
679 if ($installpid == 0)
681 close($install_stderr_r);
682 fcntl($install_stderr_w, F_SETFL, 0) or croak($!); # no close-on-exec
683 open(STDOUT, ">&", $install_stderr_w);
684 open(STDERR, ">&", $install_stderr_w);
685 srun (\@srunargs, \@execargs, {}, $build_script . $git_archive);
688 close($install_stderr_w);
689 # Tell freeze_if_want_freeze how to kill the child, otherwise the
690 # "waitpid(installpid)" loop won't get interrupted by a freeze:
691 $proc{$installpid} = {};
693 # Track whether anything appears on stderr other than slurm errors
694 # ("srun: ...") and the "starting: ..." message printed by the
695 # srun subroutine itself:
696 my $stderr_anything_from_script = 0;
697 my $match_our_own_errors = '^(srun: error: |starting: \[)';
698 while ($installpid != waitpid(-1, WNOHANG)) {
699 freeze_if_want_freeze ($installpid);
700 # Wait up to 0.1 seconds for something to appear on stderr, then
701 # do a non-blocking read.
702 my $bits = fhbits($install_stderr_r);
703 select ($bits, undef, $bits, 0.1);
704 if (0 < sysread ($install_stderr_r, $stderr_buf, 8192, length($stderr_buf)))
706 while ($stderr_buf =~ /^(.*?)\n/) {
708 substr $stderr_buf, 0, 1+length($line), "";
709 Log(undef, "stderr $line");
710 if ($line !~ /$match_our_own_errors/) {
711 $stderr_anything_from_script = 1;
716 delete $proc{$installpid};
717 $install_exited = $?;
718 close($install_stderr_r);
719 if (length($stderr_buf) > 0) {
720 if ($stderr_buf !~ /$match_our_own_errors/) {
721 $stderr_anything_from_script = 1;
723 Log(undef, "stderr $stderr_buf")
726 Log (undef, "Install script exited ".exit_status_s($install_exited));
727 last if $install_exited == 0 || $main::please_freeze;
728 # If the install script fails but doesn't print an error message,
729 # the next thing anyone is likely to do is just run it again in
730 # case it was a transient problem like "slurm communication fails
731 # because the network isn't reliable enough". So we'll just do
732 # that ourselves (up to 3 attempts in total). OTOH, if there is an
733 # error message, the problem is more likely to have a real fix and
734 # we should fail the job so the fixing process can start, instead
735 # of doing 2 more attempts.
736 last if $stderr_anything_from_script;
739 foreach my $tar_filename (map { tar_filename_n($_); } (1..$git_tar_count)) {
740 unlink($tar_filename);
743 if ($install_exited != 0) {
748 foreach (qw (script script_version script_parameters runtime_constraints))
752 (ref($Job->{$_}) ? JSON::encode_json($Job->{$_}) : $Job->{$_}));
754 foreach (split (/\n/, $Job->{knobs}))
756 Log (undef, "knob " . $_);
761 $main::success = undef;
767 my $thisround_succeeded = 0;
768 my $thisround_failed = 0;
769 my $thisround_failed_multiple = 0;
770 my $working_slot_count = scalar(@slot);
772 @jobstep_todo = sort { $jobstep[$a]->{level} <=> $jobstep[$b]->{level}
773 or $a <=> $b } @jobstep_todo;
774 my $level = $jobstep[$jobstep_todo[0]]->{level};
776 my $initial_tasks_this_level = 0;
777 foreach my $id (@jobstep_todo) {
778 $initial_tasks_this_level++ if ($jobstep[$id]->{level} == $level);
781 # If the number of tasks scheduled at this level #T is smaller than the number
782 # of slots available #S, only use the first #T slots, or the first slot on
783 # each node, whichever number is greater.
785 # When we dispatch tasks later, we'll allocate whole-node resources like RAM
786 # based on these numbers. Using fewer slots makes more resources available
787 # to each individual task, which should normally be a better strategy when
788 # there are fewer of them running with less parallelism.
790 # Note that this calculation is not redone if the initial tasks at
791 # this level queue more tasks at the same level. This may harm
792 # overall task throughput for that level.
794 if ($initial_tasks_this_level < @node) {
795 @freeslot = (0..$#node);
796 } elsif ($initial_tasks_this_level < @slot) {
797 @freeslot = (0..$initial_tasks_this_level - 1);
799 @freeslot = (0..$#slot);
801 my $round_num_freeslots = scalar(@freeslot);
803 my %round_max_slots = ();
804 for (my $ii = $#freeslot; $ii >= 0; $ii--) {
805 my $this_slot = $slot[$freeslot[$ii]];
806 my $node_name = $this_slot->{node}->{name};
807 $round_max_slots{$node_name} ||= $this_slot->{cpu};
808 last if (scalar(keys(%round_max_slots)) >= @node);
811 Log(undef, "start level $level with $round_num_freeslots slots");
814 my $progress_is_dirty = 1;
815 my $progress_stats_updated = 0;
817 update_progress_stats();
821 for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
823 my $id = $jobstep_todo[$todo_ptr];
824 my $Jobstep = $jobstep[$id];
825 if ($Jobstep->{level} != $level)
830 pipe $reader{$id}, "writer" or croak("pipe() failed: $!");
831 set_nonblocking($reader{$id});
833 my $childslot = $freeslot[0];
834 my $childnode = $slot[$childslot]->{node};
835 my $childslotname = join (".",
836 $slot[$childslot]->{node}->{name},
837 $slot[$childslot]->{cpu});
839 my $childpid = fork();
842 $SIG{'INT'} = 'DEFAULT';
843 $SIG{'QUIT'} = 'DEFAULT';
844 $SIG{'TERM'} = 'DEFAULT';
846 foreach (values (%reader))
850 fcntl ("writer", F_SETFL, 0) or croak ($!); # no close-on-exec
851 open(STDOUT,">&writer");
852 open(STDERR,">&writer");
857 delete $ENV{"GNUPGHOME"};
858 $ENV{"TASK_UUID"} = $Jobstep->{'arvados_task'}->{'uuid'};
859 $ENV{"TASK_QSEQUENCE"} = $id;
860 $ENV{"TASK_SEQUENCE"} = $level;
861 $ENV{"JOB_SCRIPT"} = $Job->{script};
862 while (my ($param, $value) = each %{$Job->{script_parameters}}) {
863 $param =~ tr/a-z/A-Z/;
864 $ENV{"JOB_PARAMETER_$param"} = $value;
866 $ENV{"TASK_SLOT_NODE"} = $slot[$childslot]->{node}->{name};
867 $ENV{"TASK_SLOT_NUMBER"} = $slot[$childslot]->{cpu};
868 $ENV{"TASK_WORK"} = $ENV{"CRUNCH_TMP"}."/task/$childslotname";
869 $ENV{"HOME"} = $ENV{"TASK_WORK"};
870 $ENV{"TASK_KEEPMOUNT"} = $ENV{"TASK_WORK"}.".keep";
871 $ENV{"TASK_TMPDIR"} = $ENV{"TASK_WORK"}; # deprecated
872 $ENV{"CRUNCH_NODE_SLOTS"} = $round_max_slots{$ENV{TASK_SLOT_NODE}};
873 $ENV{"PATH"} = $ENV{"CRUNCH_INSTALL"} . "/bin:" . $ENV{"PATH"};
879 "--nodelist=".$childnode->{name},
880 qw(-n1 -c1 -N1 -D), $ENV{'TMPDIR'},
881 "--job-name=$job_id.$id.$$",
884 my $stdbuf = " stdbuf --output=0 --error=0 ";
887 "if [ -e $ENV{TASK_WORK} ]; then rm -rf $ENV{TASK_WORK}; fi; "
888 ."mkdir -p $ENV{CRUNCH_TMP} $ENV{JOB_WORK} $ENV{TASK_WORK} $ENV{TASK_KEEPMOUNT} "
889 ."&& cd $ENV{CRUNCH_TMP} "
890 # These environment variables get used explicitly later in
891 # $command. No tool is expected to read these values directly.
892 .q{&& MEM=$(awk '($1 == "MemTotal:"){print $2}' </proc/meminfo) }
893 .q{&& SWAP=$(awk '($1 == "SwapTotal:"){print $2}' </proc/meminfo) }
894 ."&& MEMLIMIT=\$(( (\$MEM * 95) / ($ENV{CRUNCH_NODE_SLOTS} * 100) )) "
895 ."&& let SWAPLIMIT=\$MEMLIMIT+\$SWAP ";
896 $command .= "&& exec arv-mount --by-id --allow-other $ENV{TASK_KEEPMOUNT} --exec ";
899 my $cidfile = "$ENV{CRUNCH_TMP}/$Jobstep->{arvados_task}->{uuid}-$Jobstep->{failures}.cid";
900 $command .= "crunchstat -cgroup-root=/sys/fs/cgroup -cgroup-parent=docker -cgroup-cid=$cidfile -poll=10000 ";
901 $command .= "$docker_bin run --rm=true --attach=stdout --attach=stderr --attach=stdin -i \Q$dockeruserarg\E --cidfile=$cidfile --sig-proxy ";
902 # We only set memory limits if Docker lets us limit both memory and swap.
903 # Memory limits alone have been supported longer, but subprocesses tend
904 # to get SIGKILL if they exceed that without any swap limit set.
905 # See #5642 for additional background.
906 if ($docker_limitmem) {
907 $command .= "--memory=\${MEMLIMIT}k --memory-swap=\${SWAPLIMIT}k ";
910 # The source tree and $destdir directory (which we have
911 # installed on the worker host) are available in the container,
912 # under the same path.
913 $command .= "--volume=\Q$ENV{CRUNCH_SRC}:$ENV{CRUNCH_SRC}:ro\E ";
914 $command .= "--volume=\Q$ENV{CRUNCH_INSTALL}:$ENV{CRUNCH_INSTALL}:ro\E ";
916 # Currently, we make arv-mount's mount point appear at /keep
917 # inside the container (instead of using the same path as the
918 # host like we do with CRUNCH_SRC and CRUNCH_INSTALL). However,
919 # crunch scripts and utilities must not rely on this. They must
920 # use $TASK_KEEPMOUNT.
921 $command .= "--volume=\Q$ENV{TASK_KEEPMOUNT}:/keep:ro\E ";
922 $ENV{TASK_KEEPMOUNT} = "/keep";
924 # TASK_WORK is almost exactly like a docker data volume: it
925 # starts out empty, is writable, and persists until no
926 # containers use it any more. We don't use --volumes-from to
927 # share it with other containers: it is only accessible to this
928 # task, and it goes away when this task stops.
930 # However, a docker data volume is writable only by root unless
931 # the mount point already happens to exist in the container with
932 # different permissions. Therefore, we [1] assume /tmp already
933 # exists in the image and is writable by the crunch user; [2]
934 # avoid putting TASK_WORK inside CRUNCH_TMP (which won't be
935 # writable if they are created by docker while setting up the
936 # other --volumes); and [3] create $TASK_WORK inside the
937 # container using $build_script.
938 $command .= "--volume=/tmp ";
939 $ENV{"TASK_WORK"} = "/tmp/crunch-job-task-work/$childslotname";
940 $ENV{"HOME"} = $ENV{"TASK_WORK"};
941 $ENV{"TASK_TMPDIR"} = $ENV{"TASK_WORK"}; # deprecated
943 # TODO: Share a single JOB_WORK volume across all task
944 # containers on a given worker node, and delete it when the job
945 # ends (and, in case that doesn't work, when the next job
948 # For now, use the same approach as TASK_WORK above.
949 $ENV{"JOB_WORK"} = "/tmp/crunch-job-work";
951 while (my ($env_key, $env_val) = each %ENV)
953 if ($env_key =~ /^(ARVADOS|CRUNCH|JOB|TASK)_/) {
954 $command .= "--env=\Q$env_key=$env_val\E ";
957 $command .= "--env=\QHOME=$ENV{HOME}\E ";
958 $command .= "\Q$docker_hash\E ";
960 if ($Job->{arvados_sdk_version}) {
962 $command .= "perl - \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E";
964 $command .= "/bin/sh -c \'mkdir -p \"$ENV{JOB_WORK}\" \"$ENV{TASK_WORK}\" && " .
965 "if which stdbuf >/dev/null ; then " .
966 " exec $stdbuf \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E ;" .
968 " exec \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E ;" .
973 $command .= "crunchstat -cgroup-root=/sys/fs/cgroup -poll=10000 ";
975 $command .= "perl - $ENV{CRUNCH_SRC}/crunch_scripts/" . $Job->{"script"};
978 my @execargs = ('bash', '-c', $command);
979 srun (\@srunargs, \@execargs, undef, $build_script);
980 # exec() failed, we assume nothing happened.
981 die "srun() failed on build script\n";
984 if (!defined $childpid)
991 $proc{$childpid} = { jobstep => $id,
994 jobstepname => "$job_id.$id.$childpid",
996 croak ("assert failed: \$slot[$childslot]->{'pid'} exists") if exists $slot[$childslot]->{pid};
997 $slot[$childslot]->{pid} = $childpid;
999 Log ($id, "job_task ".$Jobstep->{'arvados_task'}->{'uuid'});
1000 Log ($id, "child $childpid started on $childslotname");
1001 $Jobstep->{starttime} = time;
1002 $Jobstep->{node} = $childnode->{name};
1003 $Jobstep->{slotindex} = $childslot;
1004 delete $Jobstep->{stderr};
1005 delete $Jobstep->{finishtime};
1006 delete $Jobstep->{tempfail};
1008 $Jobstep->{'arvados_task'}->{started_at} = strftime "%Y-%m-%dT%H:%M:%SZ", gmtime($Jobstep->{starttime});
1009 $Jobstep->{'arvados_task'}->save;
1011 splice @jobstep_todo, $todo_ptr, 1;
1014 $progress_is_dirty = 1;
1018 ($round_num_freeslots > @freeslot && $todo_ptr+1 > $#jobstep_todo))
1020 last THISROUND if $main::please_freeze || defined($main::success);
1021 if ($main::please_info)
1023 $main::please_info = 0;
1025 create_output_collection();
1027 update_progress_stats();
1032 if (!$gotsome || ($latest_refresh + 2 < scalar time))
1034 check_refresh_wanted();
1036 update_progress_stats();
1037 select (undef, undef, undef, 0.1);
1039 elsif (time - $progress_stats_updated >= 30 || $progress_is_dirty)
1041 update_progress_stats();
1043 $working_slot_count = scalar(grep { $_->{node}->{fail_count} == 0 &&
1044 $_->{node}->{hold_count} < 4 } @slot);
1045 if (($thisround_failed_multiple >= 8 && $thisround_succeeded == 0) ||
1046 ($thisround_failed_multiple >= 16 && $thisround_failed_multiple > $thisround_succeeded))
1048 my $message = "Repeated failure rate too high ($thisround_failed_multiple/"
1049 .($thisround_failed+$thisround_succeeded)
1050 .") -- giving up on this round";
1051 Log (undef, $message);
1055 # move slots from freeslot to holdslot (or back to freeslot) if necessary
1056 for (my $i=$#freeslot; $i>=0; $i--) {
1057 if ($slot[$freeslot[$i]]->{node}->{hold_until} > scalar time) {
1058 push @holdslot, (splice @freeslot, $i, 1);
1061 for (my $i=$#holdslot; $i>=0; $i--) {
1062 if ($slot[$holdslot[$i]]->{node}->{hold_until} <= scalar time) {
1063 push @freeslot, (splice @holdslot, $i, 1);
1067 # give up if no nodes are succeeding
1068 if ($working_slot_count < 1) {
1069 Log(undef, "Every node has failed -- giving up");
1076 push @freeslot, splice @holdslot;
1077 map { $slot[$freeslot[$_]]->{node}->{losing_streak} = 0 } (0..$#freeslot);
1080 Log (undef, "wait for last ".(scalar keys %proc)." children to finish");
1083 if ($main::please_continue) {
1084 $main::please_continue = 0;
1087 $main::please_info = 0, freeze(), create_output_collection(), save_meta(1) if $main::please_info;
1089 if (!reapchildren())
1091 check_refresh_wanted();
1093 update_progress_stats();
1094 select (undef, undef, undef, 0.1);
1095 killem (keys %proc) if $main::please_freeze;
1099 update_progress_stats();
1100 freeze_if_want_freeze();
1103 if (!defined $main::success)
1105 if (!@jobstep_todo) {
1107 } elsif ($working_slot_count < 1) {
1108 save_output_collection();
1110 exit(EX_RETRY_UNLOCKED);
1111 } elsif ($thisround_succeeded == 0 &&
1112 ($thisround_failed == 0 || $thisround_failed > 4)) {
1113 my $message = "stop because $thisround_failed tasks failed and none succeeded";
1114 Log (undef, $message);
1119 goto ONELEVEL if !defined $main::success;
1122 release_allocation();
1124 my $collated_output = save_output_collection();
1125 Log (undef, "finish");
1130 if ($collated_output && $main::success) {
1131 $final_state = 'Complete';
1133 $final_state = 'Failed';
1135 $Job->update_attributes('state' => $final_state);
1137 exit (($final_state eq 'Complete') ? 0 : 1);
1141 sub update_progress_stats
1143 $progress_stats_updated = time;
1144 return if !$progress_is_dirty;
1145 my ($todo, $done, $running) = (scalar @jobstep_todo,
1146 scalar @jobstep_done,
1147 scalar keys(%proc));
1148 $Job->{'tasks_summary'} ||= {};
1149 $Job->{'tasks_summary'}->{'todo'} = $todo;
1150 $Job->{'tasks_summary'}->{'done'} = $done;
1151 $Job->{'tasks_summary'}->{'running'} = $running;
1152 $Job->update_attributes('tasks_summary' => $Job->{'tasks_summary'});
1153 Log (undef, "status: $done done, $running running, $todo todo");
1154 $progress_is_dirty = 0;
1161 my $pid = waitpid (-1, WNOHANG);
1162 return 0 if $pid <= 0;
1164 my $whatslot = ($slot[$proc{$pid}->{slot}]->{node}->{name}
1166 . $slot[$proc{$pid}->{slot}]->{cpu});
1167 my $jobstepid = $proc{$pid}->{jobstep};
1168 my $elapsed = time - $proc{$pid}->{time};
1169 my $Jobstep = $jobstep[$jobstepid];
1171 my $childstatus = $?;
1172 my $exitvalue = $childstatus >> 8;
1173 my $exitinfo = "exit ".exit_status_s($childstatus);
1174 $Jobstep->{'arvados_task'}->reload;
1175 my $task_success = $Jobstep->{'arvados_task'}->{success};
1177 Log ($jobstepid, "child $pid on $whatslot $exitinfo success=$task_success");
1179 if (!defined $task_success) {
1180 # task did not indicate one way or the other --> fail
1181 $Jobstep->{'arvados_task'}->{success} = 0;
1182 $Jobstep->{'arvados_task'}->save;
1189 $temporary_fail ||= $Jobstep->{tempfail};
1190 $temporary_fail ||= ($exitvalue == TASK_TEMPFAIL);
1192 ++$thisround_failed;
1193 ++$thisround_failed_multiple if $Jobstep->{'failures'} >= 1;
1195 # Check for signs of a failed or misconfigured node
1196 if (++$slot[$proc{$pid}->{slot}]->{node}->{losing_streak} >=
1197 2+$slot[$proc{$pid}->{slot}]->{node}->{ncpus}) {
1198 # Don't count this against jobstep failure thresholds if this
1199 # node is already suspected faulty and srun exited quickly
1200 if ($slot[$proc{$pid}->{slot}]->{node}->{hold_until} &&
1202 Log ($jobstepid, "blaming failure on suspect node " .
1203 $slot[$proc{$pid}->{slot}]->{node}->{name});
1204 $temporary_fail ||= 1;
1206 ban_node_by_slot($proc{$pid}->{slot});
1209 Log ($jobstepid, sprintf('failure (#%d, %s) after %d seconds',
1210 ++$Jobstep->{'failures'},
1211 $temporary_fail ? 'temporary' : 'permanent',
1214 if (!$temporary_fail || $Jobstep->{'failures'} >= 3) {
1215 # Give up on this task, and the whole job
1218 # Put this task back on the todo queue
1219 push @jobstep_todo, $jobstepid;
1220 $Job->{'tasks_summary'}->{'failed'}++;
1224 ++$thisround_succeeded;
1225 $slot[$proc{$pid}->{slot}]->{node}->{losing_streak} = 0;
1226 $slot[$proc{$pid}->{slot}]->{node}->{hold_until} = 0;
1227 $slot[$proc{$pid}->{slot}]->{node}->{fail_count} = 0;
1228 push @jobstep_done, $jobstepid;
1229 Log ($jobstepid, "success in $elapsed seconds");
1231 $Jobstep->{exitcode} = $childstatus;
1232 $Jobstep->{finishtime} = time;
1233 $Jobstep->{'arvados_task'}->{finished_at} = strftime "%Y-%m-%dT%H:%M:%SZ", gmtime($Jobstep->{finishtime});
1234 $Jobstep->{'arvados_task'}->save;
1235 process_stderr ($jobstepid, $task_success);
1236 Log ($jobstepid, sprintf("task output (%d bytes): %s",
1237 length($Jobstep->{'arvados_task'}->{output}),
1238 $Jobstep->{'arvados_task'}->{output}));
1240 close $reader{$jobstepid};
1241 delete $reader{$jobstepid};
1242 delete $slot[$proc{$pid}->{slot}]->{pid};
1243 push @freeslot, $proc{$pid}->{slot};
1246 if ($task_success) {
1248 my $newtask_list = [];
1249 my $newtask_results;
1251 $newtask_results = api_call(
1254 'created_by_job_task_uuid' => $Jobstep->{'arvados_task'}->{uuid}
1256 'order' => 'qsequence',
1257 'offset' => scalar(@$newtask_list),
1259 push(@$newtask_list, @{$newtask_results->{items}});
1260 } while (@{$newtask_results->{items}});
1261 foreach my $arvados_task (@$newtask_list) {
1263 'level' => $arvados_task->{'sequence'},
1265 'arvados_task' => $arvados_task
1267 push @jobstep, $jobstep;
1268 push @jobstep_todo, $#jobstep;
1272 $progress_is_dirty = 1;
1276 sub check_refresh_wanted
1278 my @stat = stat $ENV{"CRUNCH_REFRESH_TRIGGER"};
1279 if (@stat && $stat[9] > $latest_refresh) {
1280 $latest_refresh = scalar time;
1281 my $Job2 = api_call("jobs/get", uuid => $jobspec);
1282 for my $attr ('cancelled_at',
1283 'cancelled_by_user_uuid',
1284 'cancelled_by_client_uuid',
1286 $Job->{$attr} = $Job2->{$attr};
1288 if ($Job->{'state'} ne "Running") {
1289 if ($Job->{'state'} eq "Cancelled") {
1290 Log (undef, "Job cancelled at " . $Job->{'cancelled_at'} . " by user " . $Job->{'cancelled_by_user_uuid'});
1292 Log (undef, "Job state unexpectedly changed to " . $Job->{'state'});
1295 $main::please_freeze = 1;
1302 my $last_squeue_check = $squeue_checked;
1304 # Do not call `squeue` or check the kill list more than once every
1306 return if $last_squeue_check > time - 15;
1307 $squeue_checked = time;
1309 # Look for children from which we haven't received stderr data since
1310 # the last squeue check. If no such children exist, all procs are
1311 # alive and there's no need to even look at squeue.
1313 # As long as the crunchstat poll interval (10s) is shorter than the
1314 # squeue check interval (15s) this should make the squeue check an
1316 my $silent_procs = 0;
1317 for my $jobstep (values %proc)
1319 if ($jobstep->{stderr_at} < $last_squeue_check)
1324 return if $silent_procs == 0;
1326 # use killem() on procs whose killtime is reached
1327 while (my ($pid, $jobstep) = each %proc)
1329 if (exists $jobstep->{killtime}
1330 && $jobstep->{killtime} <= time
1331 && $jobstep->{stderr_at} < $last_squeue_check)
1334 if ($jobstep->{stderr_at}) {
1335 $sincewhen = " in last " . (time - $jobstep->{stderr_at}) . "s";
1337 Log($jobstep->{jobstep}, "killing orphaned srun process $pid (task not in slurm queue, no stderr received$sincewhen)");
1344 # here is an opportunity to check for mysterious problems with local procs
1348 # Get a list of steps still running. Note: squeue(1) says --steps
1349 # selects a format (which we override anyway) and allows us to
1350 # specify which steps we're interested in (which we don't).
1351 # Importantly, it also changes the meaning of %j from "job name" to
1352 # "step name" and (although this isn't mentioned explicitly in the
1353 # docs) switches from "one line per job" mode to "one line per step"
1354 # mode. Without it, we'd just get a list of one job, instead of a
1356 my @squeue = `squeue --jobs=\Q$ENV{SLURM_JOB_ID}\E --steps --format='%j' --noheader`;
1359 Log(undef, "warning: squeue exit status $? ($!)");
1364 # which of my jobsteps are running, according to squeue?
1366 for my $jobstepname (@squeue)
1368 $ok{$jobstepname} = 1;
1371 # Check for child procs >60s old and not mentioned by squeue.
1372 while (my ($pid, $jobstep) = each %proc)
1374 if ($jobstep->{time} < time - 60
1375 && $jobstep->{jobstepname}
1376 && !exists $ok{$jobstep->{jobstepname}}
1377 && !exists $jobstep->{killtime})
1379 # According to slurm, this task has ended (successfully or not)
1380 # -- but our srun child hasn't exited. First we must wait (30
1381 # seconds) in case this is just a race between communication
1382 # channels. Then, if our srun child process still hasn't
1383 # terminated, we'll conclude some slurm communication
1384 # error/delay has caused the task to die without notifying srun,
1385 # and we'll kill srun ourselves.
1386 $jobstep->{killtime} = time + 30;
1387 Log($jobstep->{jobstep}, "notice: task is not in slurm queue but srun process $pid has not exited");
1393 sub release_allocation
1397 Log (undef, "release job allocation");
1398 system "scancel $ENV{SLURM_JOB_ID}";
1406 foreach my $job (keys %reader)
1409 while (0 < sysread ($reader{$job}, $buf, 8192))
1411 print STDERR $buf if $ENV{CRUNCH_DEBUG};
1412 $jobstep[$job]->{stderr_at} = time;
1413 $jobstep[$job]->{stderr} .= $buf;
1414 preprocess_stderr ($job);
1415 if (length ($jobstep[$job]->{stderr}) > 16384)
1417 substr ($jobstep[$job]->{stderr}, 0, 8192) = "";
1426 sub preprocess_stderr
1430 while ($jobstep[$job]->{stderr} =~ /^(.*?)\n/) {
1432 substr $jobstep[$job]->{stderr}, 0, 1+length($line), "";
1433 Log ($job, "stderr $line");
1434 if ($line =~ /srun: error: (SLURM job $ENV{SLURM_JOB_ID} has expired|Unable to confirm allocation for job $ENV{SLURM_JOB_ID})/) {
1436 $main::please_freeze = 1;
1438 elsif ($line =~ /srun: error: Node failure on/) {
1439 my $job_slot_index = $jobstep[$job]->{slotindex};
1440 $slot[$job_slot_index]->{node}->{fail_count}++;
1441 $jobstep[$job]->{tempfail} = 1;
1442 ban_node_by_slot($job_slot_index);
1444 elsif ($line =~ /srun: error: (Unable to create job step|.*: Communication connection failure)/) {
1445 $jobstep[$job]->{tempfail} = 1;
1446 ban_node_by_slot($jobstep[$job]->{slotindex});
1448 elsif ($line =~ /arvados\.errors\.Keep/) {
1449 $jobstep[$job]->{tempfail} = 1;
1458 my $task_success = shift;
1459 preprocess_stderr ($job);
1462 Log ($job, "stderr $_");
1463 } split ("\n", $jobstep[$job]->{stderr});
1470 if (!open($keep, "-|", "arv-get", "--retries", retry_count(), $hash)) {
1471 Log(undef, "fetch_block run error from arv-get $hash: $!");
1474 my $output_block = "";
1477 my $bytes = sysread($keep, $buf, 1024 * 1024);
1478 if (!defined $bytes) {
1479 Log(undef, "fetch_block read error from arv-get: $!");
1480 $output_block = undef;
1482 } elsif ($bytes == 0) {
1483 # sysread returns 0 at the end of the pipe.
1486 # some bytes were read into buf.
1487 $output_block .= $buf;
1492 Log(undef, "fetch_block arv-get exited " . exit_status_s($?));
1493 $output_block = undef;
1495 return $output_block;
1498 # Create a collection by concatenating the output of all tasks (each
1499 # task's output is either a manifest fragment, a locator for a
1500 # manifest fragment stored in Keep, or nothing at all). Return the
1501 # portable_data_hash of the new collection.
1502 sub create_output_collection
1504 Log (undef, "collate");
1506 my ($child_out, $child_in);
1507 my $pid = open2($child_out, $child_in, 'python', '-c', q{
1510 print (arvados.api("v1").collections().
1511 create(body={"manifest_text": sys.stdin.read()}).
1512 execute(num_retries=int(sys.argv[1]))["portable_data_hash"])
1516 my $manifest_size = 0;
1520 my $output = $_->{'arvados_task'}->{output};
1521 next if (!defined($output));
1523 if ($output =~ /^[0-9a-f]{32}(\+\S+)*$/) {
1524 $next_write = fetch_block($output);
1526 $next_write = $output;
1528 if (defined($next_write)) {
1529 if (!defined(syswrite($child_in, $next_write))) {
1530 # There's been an error writing. Stop the loop.
1531 # We'll log details about the exit code later.
1534 $manifest_size += length($next_write);
1537 my $uuid = $_->{'arvados_task'}->{'uuid'};
1538 Log (undef, "Error retrieving '$output' output by task $task_idx ($uuid)");
1543 Log(undef, "collated output manifest text to send to API server is $manifest_size bytes with access tokens");
1546 my $s = IO::Select->new($child_out);
1547 if ($s->can_read(120)) {
1548 sysread($child_out, $joboutput, 1024 * 1024);
1551 Log(undef, "output collection creation exited " . exit_status_s($?));
1557 Log (undef, "timed out while creating output collection");
1558 foreach my $signal (2, 2, 2, 15, 15, 9) {
1559 kill($signal, $pid);
1560 last if waitpid($pid, WNOHANG) == -1;
1569 # Calls create_output_collection, logs the result, and returns it.
1570 # If that was successful, save that as the output in the job record.
1571 sub save_output_collection {
1572 my $collated_output = create_output_collection();
1574 if (!$collated_output) {
1575 Log(undef, "Failed to write output collection");
1578 Log(undef, "job output $collated_output");
1579 $Job->update_attributes('output' => $collated_output);
1581 return $collated_output;
1588 my $sig = 2; # SIGINT first
1589 if (exists $proc{$_}->{"sent_$sig"} &&
1590 time - $proc{$_}->{"sent_$sig"} > 4)
1592 $sig = 15; # SIGTERM if SIGINT doesn't work
1594 if (exists $proc{$_}->{"sent_$sig"} &&
1595 time - $proc{$_}->{"sent_$sig"} > 4)
1597 $sig = 9; # SIGKILL if SIGTERM doesn't work
1599 if (!exists $proc{$_}->{"sent_$sig"})
1601 Log ($proc{$_}->{jobstep}, "sending 2x signal $sig to pid $_");
1603 select (undef, undef, undef, 0.1);
1606 kill $sig, $_; # srun wants two SIGINT to really interrupt
1608 $proc{$_}->{"sent_$sig"} = time;
1609 $proc{$_}->{"killedafter"} = time - $proc{$_}->{"time"};
1619 vec($bits,fileno($_),1) = 1;
1625 # Send log output to Keep via arv-put.
1627 # $log_pipe_in and $log_pipe_out are the input and output filehandles to the arv-put pipe.
1628 # $log_pipe_out_buf is a string containing all output read from arv-put so far.
1629 # $log_pipe_out_select is an IO::Select object around $log_pipe_out.
1630 # $log_pipe_pid is the pid of the arv-put subprocess.
1632 # The only functions that should access these variables directly are:
1634 # log_writer_start($logfilename)
1635 # Starts an arv-put pipe, reading data on stdin and writing it to
1636 # a $logfilename file in an output collection.
1638 # log_writer_read_output([$timeout])
1639 # Read output from $log_pipe_out and append it to $log_pipe_out_buf.
1640 # Passes $timeout to the select() call, with a default of 0.01.
1641 # Returns the result of the last read() call on $log_pipe_out, or
1642 # -1 if read() wasn't called because select() timed out.
1643 # Only other log_writer_* functions should need to call this.
1645 # log_writer_send($txt)
1646 # Writes $txt to the output log collection.
1648 # log_writer_finish()
1649 # Closes the arv-put pipe and returns the output that it produces.
1651 # log_writer_is_active()
1652 # Returns a true value if there is currently a live arv-put
1653 # process, false otherwise.
1655 my ($log_pipe_in, $log_pipe_out, $log_pipe_out_buf, $log_pipe_out_select,
1658 sub log_writer_start($)
1660 my $logfilename = shift;
1661 $log_pipe_pid = open2($log_pipe_out, $log_pipe_in,
1665 '--filename', $logfilename,
1667 $log_pipe_out_buf = "";
1668 $log_pipe_out_select = IO::Select->new($log_pipe_out);
1671 sub log_writer_read_output {
1672 my $timeout = shift || 0.01;
1674 while ($read && $log_pipe_out_select->can_read($timeout)) {
1675 $read = read($log_pipe_out, $log_pipe_out_buf, 65536,
1676 length($log_pipe_out_buf));
1678 if (!defined($read)) {
1679 Log(undef, "error reading log manifest from arv-put: $!");
1684 sub log_writer_send($)
1687 print $log_pipe_in $txt;
1688 log_writer_read_output();
1691 sub log_writer_finish()
1693 return unless $log_pipe_pid;
1695 close($log_pipe_in);
1697 my $read_result = log_writer_read_output(120);
1698 if ($read_result == -1) {
1699 Log (undef, "timed out reading from 'arv-put'");
1700 } elsif ($read_result != 0) {
1701 Log(undef, "failed to read arv-put log manifest to EOF");
1704 waitpid($log_pipe_pid, 0);
1706 Log(undef, "log_writer_finish: arv-put exited " . exit_status_s($?))
1709 close($log_pipe_out);
1710 my $arv_put_output = $log_pipe_out_buf;
1711 $log_pipe_pid = $log_pipe_in = $log_pipe_out = $log_pipe_out_buf =
1712 $log_pipe_out_select = undef;
1714 return $arv_put_output;
1717 sub log_writer_is_active() {
1718 return $log_pipe_pid;
1721 sub Log # ($jobstep_id, $logmessage)
1723 if ($_[1] =~ /\n/) {
1724 for my $line (split (/\n/, $_[1])) {
1729 my $fh = select STDERR; $|=1; select $fh;
1730 my $message = sprintf ("%s %d %s %s", $job_id, $$, @_);
1731 $message =~ s{([^ -\176])}{"\\" . sprintf ("%03o", ord($1))}ge;
1734 if (log_writer_is_active() || -t STDERR) {
1735 my @gmtime = gmtime;
1736 $datetime = sprintf ("%04d-%02d-%02d_%02d:%02d:%02d",
1737 $gmtime[5]+1900, $gmtime[4]+1, @gmtime[3,2,1,0]);
1739 print STDERR ((-t STDERR) ? ($datetime." ".$message) : $message);
1741 if (log_writer_is_active()) {
1742 log_writer_send($datetime . " " . $message);
1749 my ($package, $file, $line) = caller;
1750 my $message = "@_ at $file line $line\n";
1751 Log (undef, $message);
1752 freeze() if @jobstep_todo;
1753 create_output_collection() if @jobstep_todo;
1763 if ($Job->{'state'} eq 'Cancelled') {
1764 $Job->update_attributes('finished_at' => scalar gmtime);
1766 $Job->update_attributes('state' => 'Failed');
1773 my $justcheckpoint = shift; # false if this will be the last meta saved
1774 return if $justcheckpoint; # checkpointing is not relevant post-Warehouse.pm
1775 return unless log_writer_is_active();
1777 my $log_manifest = "";
1779 my $prev_log_coll = api_call("collections/get", uuid => $Job->{log});
1780 $log_manifest .= $prev_log_coll->{manifest_text};
1782 $log_manifest .= log_writer_finish();
1784 my $log_coll = api_call(
1785 "collections/create", ensure_unique_name => 1, collection => {
1786 manifest_text => $log_manifest,
1787 owner_uuid => $Job->{owner_uuid},
1788 name => sprintf("Log from %s job %s", $Job->{script}, $Job->{uuid}),
1790 Log(undef, "log collection is " . $log_coll->{portable_data_hash});
1791 $Job->update_attributes('log' => $log_coll->{portable_data_hash});
1795 sub freeze_if_want_freeze
1797 if ($main::please_freeze)
1799 release_allocation();
1802 # kill some srun procs before freeze+stop
1803 map { $proc{$_} = {} } @_;
1806 killem (keys %proc);
1807 select (undef, undef, undef, 0.1);
1809 while (($died = waitpid (-1, WNOHANG)) > 0)
1811 delete $proc{$died};
1816 create_output_collection();
1826 Log (undef, "Freeze not implemented");
1833 croak ("Thaw not implemented");
1849 $s =~ s{\\(.)}{$1 eq "n" ? "\n" : $1}ge;
1856 my $srunargs = shift;
1857 my $execargs = shift;
1858 my $opts = shift || {};
1860 my $args = $have_slurm ? [@$srunargs, @$execargs] : $execargs;
1862 $Data::Dumper::Terse = 1;
1863 $Data::Dumper::Indent = 0;
1864 my $show_cmd = Dumper($args);
1865 $show_cmd =~ s/(TOKEN\\*=)[^\s\']+/${1}[...]/g;
1866 $show_cmd =~ s/\n/ /g;
1867 if ($opts->{fork}) {
1868 Log(undef, "starting: $show_cmd");
1870 # This is a child process: parent is in charge of reading our
1871 # stderr and copying it to Log() if needed.
1872 warn "starting: $show_cmd\n";
1875 if (defined $stdin) {
1876 my $child = open STDIN, "-|";
1877 defined $child or die "no fork: $!";
1879 print $stdin or die $!;
1880 close STDOUT or die $!;
1885 return system (@$args) if $opts->{fork};
1888 warn "ENV size is ".length(join(" ",%ENV));
1889 die "exec failed: $!: @$args";
1893 sub ban_node_by_slot {
1894 # Don't start any new jobsteps on this node for 60 seconds
1896 $slot[$slotid]->{node}->{hold_until} = 60 + scalar time;
1897 $slot[$slotid]->{node}->{hold_count}++;
1898 Log (undef, "backing off node " . $slot[$slotid]->{node}->{name} . " for 60 seconds");
1903 my ($lockfile, $error_message) = @_;
1904 open L, ">", $lockfile or croak("$lockfile: $!");
1905 if (!flock L, LOCK_EX|LOCK_NB) {
1906 croak("Can't lock $lockfile: $error_message\n");
1910 sub find_docker_image {
1911 # Given a Keep locator, check to see if it contains a Docker image.
1912 # If so, return its stream name and Docker hash.
1913 # If not, return undef for both values.
1914 my $locator = shift;
1915 my ($streamname, $filename);
1916 my $image = api_call("collections/get", uuid => $locator);
1918 foreach my $line (split(/\n/, $image->{manifest_text})) {
1919 my @tokens = split(/\s+/, $line);
1921 $streamname = shift(@tokens);
1922 foreach my $filedata (grep(/^\d+:\d+:/, @tokens)) {
1923 if (defined($filename)) {
1924 return (undef, undef); # More than one file in the Collection.
1926 $filename = (split(/:/, $filedata, 3))[2];
1931 if (defined($filename) and ($filename =~ /^([0-9A-Fa-f]{64})\.tar$/)) {
1932 return ($streamname, $1);
1934 return (undef, undef);
1939 # Calculate the number of times an operation should be retried,
1940 # assuming exponential backoff, and that we're willing to retry as
1941 # long as tasks have been running. Enforce a minimum of 3 retries.
1942 my ($starttime, $endtime, $timediff, $retries);
1944 $starttime = $jobstep[0]->{starttime};
1945 $endtime = $jobstep[-1]->{finishtime};
1947 if (!defined($starttime)) {
1949 } elsif (!defined($endtime)) {
1950 $timediff = time - $starttime;
1952 $timediff = ($endtime - $starttime) - (time - $endtime);
1954 if ($timediff > 0) {
1955 $retries = int(log($timediff) / log(2));
1957 $retries = 1; # Use the minimum.
1959 return ($retries > 3) ? $retries : 3;
1963 # Pass in two function references.
1964 # This method will be called with the remaining arguments.
1965 # If it dies, retry it with exponential backoff until it succeeds,
1966 # or until the current retry_count is exhausted. After each failure
1967 # that can be retried, the second function will be called with
1968 # the current try count (0-based), next try time, and error message.
1969 my $operation = shift;
1970 my $retry_callback = shift;
1971 my $retries = retry_count();
1972 foreach my $try_count (0..$retries) {
1973 my $next_try = time + (2 ** $try_count);
1974 my $result = eval { $operation->(@_); };
1977 } elsif ($try_count < $retries) {
1978 $retry_callback->($try_count, $next_try, $@);
1979 my $sleep_time = $next_try - time;
1980 sleep($sleep_time) if ($sleep_time > 0);
1983 # Ensure the error message ends in a newline, so Perl doesn't add
1984 # retry_op's line number to it.
1990 # Pass in a /-separated API method name, and arguments for it.
1991 # This function will call that method, retrying as needed until
1992 # the current retry_count is exhausted, with a log on the first failure.
1993 my $method_name = shift;
1994 my $log_api_retry = sub {
1995 my ($try_count, $next_try_at, $errmsg) = @_;
1996 $errmsg =~ s/\s*\bat \Q$0\E line \d+\.?\s*//;
1997 $errmsg =~ s/\s/ /g;
1998 $errmsg =~ s/\s+$//;
2000 if ($next_try_at < time) {
2001 $retry_msg = "Retrying.";
2003 my $next_try_fmt = strftime "%Y-%m-%dT%H:%M:%SZ", gmtime($next_try_at);
2004 $retry_msg = "Retrying at $next_try_fmt.";
2006 Log(undef, "API method $method_name failed: $errmsg. $retry_msg");
2009 foreach my $key (split(/\//, $method_name)) {
2010 $method = $method->{$key};
2012 return retry_op(sub { $method->execute(@_); }, $log_api_retry, @_);
2016 # Given a $?, return a human-readable exit code string like "0" or
2017 # "1" or "0 with signal 1" or "1 with signal 11".
2018 my $exitcode = shift;
2019 my $s = $exitcode >> 8;
2020 if ($exitcode & 0x7f) {
2021 $s .= " with signal " . ($exitcode & 0x7f);
2023 if ($exitcode & 0x80) {
2024 $s .= " with core dump";
2029 sub handle_readall {
2030 # Pass in a glob reference to a file handle.
2031 # Read all its contents and return them as a string.
2032 my $fh_glob_ref = shift;
2034 return <$fh_glob_ref>;
2037 sub tar_filename_n {
2039 return sprintf("%s/git.%s.%d.tar", $ENV{CRUNCH_TMP}, $job_id, $n);
2042 sub add_git_archive {
2043 # Pass in a git archive command as a string or list, a la system().
2044 # This method will save its output to be included in the archive sent to the
2048 if (!open(GIT_ARCHIVE, ">", tar_filename_n($git_tar_count))) {
2049 croak("Failed to save git archive: $!");
2051 my $git_pid = open2(">&GIT_ARCHIVE", $git_input, @_);
2053 waitpid($git_pid, 0);
2056 croak("Failed to save git archive: git exited " . exit_status_s($?));
2060 sub combined_git_archive {
2061 # Combine all saved tar archives into a single archive, then return its
2062 # contents in a string. Return undef if no archives have been saved.
2063 if ($git_tar_count < 1) {
2066 my $base_tar_name = tar_filename_n(1);
2067 foreach my $tar_to_append (map { tar_filename_n($_); } (2..$git_tar_count)) {
2068 my $tar_exit = system("tar", "-Af", $base_tar_name, $tar_to_append);
2069 if ($tar_exit != 0) {
2070 croak("Error preparing build archive: tar -A exited " .
2071 exit_status_s($tar_exit));
2074 if (!open(GIT_TAR, "<", $base_tar_name)) {
2075 croak("Could not open build archive: $!");
2077 my $tar_contents = handle_readall(\*GIT_TAR);
2079 return $tar_contents;
2082 sub set_nonblocking {
2084 my $flags = fcntl ($fh, F_GETFL, 0) or croak ($!);
2085 fcntl ($fh, F_SETFL, $flags | O_NONBLOCK) or croak ($!);
2091 # This is crunch-job's internal dispatch script. crunch-job running on the API
2092 # server invokes this script on individual compute nodes, or localhost if we're
2093 # running a job locally. It gets called in two modes:
2095 # * No arguments: Installation mode. Read a tar archive from the DATA
2096 # file handle; it includes the Crunch script's source code, and
2097 # maybe SDKs as well. Those should be installed in the proper
2098 # locations. This runs outside of any Docker container, so don't try to
2099 # introspect Crunch's runtime environment.
2101 # * With arguments: Crunch script run mode. This script should set up the
2102 # environment, then run the command specified in the arguments. This runs
2103 # inside any Docker container.
2106 use File::Path qw( make_path remove_tree );
2107 use POSIX qw(getcwd);
2109 use constant TASK_TEMPFAIL => 111;
2111 # Map SDK subdirectories to the path environments they belong to.
2112 my %SDK_ENVVARS = ("perl/lib" => "PERLLIB", "ruby/lib" => "RUBYLIB");
2114 my $destdir = $ENV{"CRUNCH_SRC"};
2115 my $archive_hash = $ENV{"CRUNCH_GIT_ARCHIVE_HASH"};
2116 my $repo = $ENV{"CRUNCH_SRC_URL"};
2117 my $install_dir = $ENV{"CRUNCH_INSTALL"} || (getcwd() . "/opt");
2118 my $job_work = $ENV{"JOB_WORK"};
2119 my $task_work = $ENV{"TASK_WORK"};
2121 open(STDOUT_ORIG, ">&", STDOUT);
2122 open(STDERR_ORIG, ">&", STDERR);
2124 for my $dir ($destdir, $job_work, $task_work) {
2127 -e $dir or die "Failed to create temporary directory ($dir): $!";
2132 remove_tree($task_work, {keep_root => 1});
2135 ### Crunch script run mode
2137 # We want to do routine logging during task 0 only. This gives the user
2138 # the information they need, but avoids repeating the information for every
2141 if ($ENV{TASK_SEQUENCE} eq "0") {
2144 printf STDERR_ORIG "[Crunch] $msg\n", @_;
2150 my $python_src = "$install_dir/python";
2151 my $venv_dir = "$job_work/.arvados.venv";
2152 my $venv_built = -e "$venv_dir/bin/activate";
2153 if ((!$venv_built) and (-d $python_src) and can_run("virtualenv")) {
2154 shell_or_die(undef, "virtualenv", "--quiet", "--system-site-packages",
2155 "--python=python2.7", $venv_dir);
2156 shell_or_die(TASK_TEMPFAIL, "$venv_dir/bin/pip", "--quiet", "install", "-I", $python_src);
2158 $Log->("Built Python SDK virtualenv");
2161 my $pip_bin = "pip";
2163 $Log->("Running in Python SDK virtualenv");
2164 $pip_bin = "$venv_dir/bin/pip";
2165 my $orig_argv = join(" ", map { quotemeta($_); } @ARGV);
2166 @ARGV = ("/bin/sh", "-ec",
2167 ". \Q$venv_dir/bin/activate\E; exec $orig_argv");
2168 } elsif (-d $python_src) {
2169 $Log->("Warning: virtualenv not found inside Docker container default " .
2170 "\$PATH. Can't install Python SDK.");
2173 my $pkgs = `(\Q$pip_bin\E freeze 2>/dev/null | grep arvados) || dpkg-query --show '*arvados*'`;
2175 $Log->("Using Arvados SDK:");
2176 foreach my $line (split /\n/, $pkgs) {
2180 $Log->("Arvados SDK packages not found");
2183 while (my ($sdk_dir, $sdk_envkey) = each(%SDK_ENVVARS)) {
2184 my $sdk_path = "$install_dir/$sdk_dir";
2186 if ($ENV{$sdk_envkey}) {
2187 $ENV{$sdk_envkey} = "$sdk_path:" . $ENV{$sdk_envkey};
2189 $ENV{$sdk_envkey} = $sdk_path;
2191 $Log->("Arvados SDK added to %s", $sdk_envkey);
2196 die "Cannot exec `@ARGV`: $!";
2199 ### Installation mode
2200 open L, ">", "$destdir.lock" or die "$destdir.lock: $!";
2202 if (readlink ("$destdir.archive_hash") eq $archive_hash && -d $destdir) {
2203 # This exact git archive (source + arvados sdk) is already installed
2204 # here, so there's no need to reinstall it.
2206 # We must consume our DATA section, though: otherwise the process
2207 # feeding it to us will get SIGPIPE.
2209 while (read(DATA, $buf, 65536)) { }
2214 unlink "$destdir.archive_hash";
2218 # Ignore SIGPIPE: we check retval of close() instead. See perlipc(1).
2219 local $SIG{PIPE} = "IGNORE";
2220 warn "Extracting archive: $archive_hash\n";
2221 # --ignore-zeros is necessary sometimes: depending on how much NUL
2222 # padding tar -A put on our combined archive (which in turn depends
2223 # on the length of the component archives) tar without
2224 # --ignore-zeros will exit before consuming stdin and cause close()
2225 # to fail on the resulting SIGPIPE.
2226 if (!open(TARX, "|-", "tar", "--ignore-zeros", "-xC", $destdir)) {
2227 die "Error launching 'tar -xC $destdir': $!";
2229 # If we send too much data to tar in one write (> 4-5 MiB), it stops, and we
2230 # get SIGPIPE. We must feed it data incrementally.
2232 while (read(DATA, $tar_input, 65536)) {
2233 print TARX $tar_input;
2236 die "'tar -xC $destdir' exited $?: $!";
2242 my $sdk_root = "$destdir/.arvados.sdk/sdk";
2244 foreach my $sdk_lang (("python",
2245 map { (split /\//, $_, 2)[0]; } keys(%SDK_ENVVARS))) {
2246 if (-d "$sdk_root/$sdk_lang") {
2247 if (!rename("$sdk_root/$sdk_lang", "$install_dir/$sdk_lang")) {
2248 die "Failed to install $sdk_lang SDK: $!";
2254 my $python_dir = "$install_dir/python";
2255 if ((-d $python_dir) and can_run("python2.7")) {
2256 open(my $egg_info_pipe, "-|",
2257 "python2.7 \Q$python_dir/setup.py\E egg_info 2>&1 >/dev/null");
2258 my @egg_info_errors = <$egg_info_pipe>;
2259 close($egg_info_pipe);
2262 if (@egg_info_errors and (($egg_info_errors[-1] =~ /\bgit\b/) or ($egg_info_errors[-1] =~ /\[Errno 2\]/))) {
2263 # egg_info apparently failed because it couldn't ask git for a build tag.
2264 # Specify no build tag.
2265 open(my $pysdk_cfg, ">>", "$python_dir/setup.cfg");
2266 print $pysdk_cfg "\n[egg_info]\ntag_build =\n";
2269 my $egg_info_exit = $? >> 8;
2270 foreach my $errline (@egg_info_errors) {
2273 warn "python setup.py egg_info failed: exit $egg_info_exit";
2274 exit ($egg_info_exit || 1);
2279 # Hide messages from the install script (unless it fails: shell_or_die
2280 # will show $destdir.log in that case).
2281 open(STDOUT, ">>", "$destdir.log");
2282 open(STDERR, ">&", STDOUT);
2284 if (-e "$destdir/crunch_scripts/install") {
2285 shell_or_die (undef, "$destdir/crunch_scripts/install", $install_dir);
2286 } elsif (!-e "./install.sh" && -e "./tests/autotests.sh") {
2288 shell_or_die (undef, "./tests/autotests.sh", $install_dir);
2289 } elsif (-e "./install.sh") {
2290 shell_or_die (undef, "./install.sh", $install_dir);
2293 if ($archive_hash) {
2294 unlink "$destdir.archive_hash.new";
2295 symlink ($archive_hash, "$destdir.archive_hash.new") or die "$destdir.archive_hash.new: $!";
2296 rename ("$destdir.archive_hash.new", "$destdir.archive_hash") or die "$destdir.archive_hash: $!";
2302 my $command_name = shift;
2303 open(my $which, "-|", "which", $command_name);
2304 while (<$which>) { }
2311 my $exitcode = shift;
2313 if ($ENV{"DEBUG"}) {
2314 print STDERR "@_\n";
2316 if (system (@_) != 0) {
2319 my $exitstatus = sprintf("exit %d signal %d", $code >> 8, $code & 0x7f);
2320 open STDERR, ">&STDERR_ORIG";
2321 system ("cat $destdir.log >&2");
2322 warn "@_ failed ($err): $exitstatus";
2323 if (defined($exitcode)) {
2327 exit (($code >> 8) || 1);