X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/f16be1736d705278ae39bde3fc6d6d9d1f302fa7..04bce28b705a6ca33d2c9b19aa783f588a99c8e0:/sdk/cli/bin/crunch-job diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job index 47f9894a41..0d31afcb87 100755 --- a/sdk/cli/bin/crunch-job +++ b/sdk/cli/bin/crunch-job @@ -394,7 +394,7 @@ else if ($cleanpid == 0) { srun (["srun", "--nodelist=$nodelist", "-D", $ENV{'TMPDIR'}], - ['bash', '-c', 'if mount | grep -q $JOB_WORK/; then sudo /bin/umount $JOB_WORK/* 2>/dev/null; fi; sleep 1; rm -rf $JOB_WORK $CRUNCH_TMP/opt $CRUNCH_TMP/src*']); + ['bash', '-c', 'if mount | grep -q $JOB_WORK/; then for i in $JOB_WORK/*keep; do /bin/fusermount -z -u $i; done; fi; sleep 1; rm -rf $JOB_WORK $CRUNCH_TMP/opt $CRUNCH_TMP/src*']); exit (1); } while (1) @@ -487,6 +487,8 @@ else croak ("could not figure out commit id for $treeish"); } + # Note: this section is almost certainly unnecessary if we're + # running tasks in docker containers. my $installpid = fork(); if ($installpid == 0) { @@ -649,13 +651,11 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++) my $build_script_to_send = ""; my $command = "if [ -e $ENV{TASK_WORK} ]; then rm -rf $ENV{TASK_WORK}; fi; " - ."mkdir -p $ENV{JOB_WORK} $ENV{CRUNCH_TMP} $ENV{TASK_WORK} $ENV{TASK_KEEPMOUNT}" + ."mkdir -p $ENV{CRUNCH_TMP} $ENV{JOB_WORK} $ENV{TASK_WORK} $ENV{TASK_KEEPMOUNT} " ."&& cd $ENV{CRUNCH_TMP} "; if ($build_script) { $build_script_to_send = $build_script; - $command .= - "&& perl -"; } $command .= "&& exec arv-mount --allow-other $ENV{TASK_KEEPMOUNT} --exec "; if ($docker_hash) @@ -668,16 +668,12 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++) $command .= q{$(ip -o address show scope global | gawk 'match($4, /^([0-9\.:]+)\//, x){print "--dns", x[1]}') }; - $command .= "--volume=\Q$ENV{CRUNCH_SRC}:/tmp/crunch-src:ro\E "; $command .= "--volume=\Q$ENV{TASK_KEEPMOUNT}:/keep:ro\E "; $command .= "--env=\QHOME=/home/crunch\E "; while (my ($env_key, $env_val) = each %ENV) { if ($env_key =~ /^(ARVADOS|JOB|TASK)_/) { - if ($env_key eq "TASK_WORK") { - $command .= "--env=\QTASK_WORK=/tmp/crunch-job\E "; - } - elsif ($env_key eq "TASK_KEEPMOUNT") { + if ($env_key eq "TASK_KEEPMOUNT") { $command .= "--env=\QTASK_KEEPMOUNT=/keep\E "; } else { @@ -689,11 +685,13 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++) $command .= "--env=\QCRUNCH_SRC=/tmp/crunch-src\E "; $command .= "\Q$docker_hash\E "; $command .= "stdbuf --output=0 --error=0 "; + $command .= "perl - " if ($build_script); $command .= "/tmp/crunch-src/crunch_scripts/" . $Job->{"script"}; } else { # Non-docker run $command .= "crunchstat -cgroup-root=/sys/fs/cgroup -poll=10000 "; $command .= "stdbuf --output=0 --error=0 "; + $command .= "perl - " if ($build_script); $command .= "$ENV{CRUNCH_SRC}/crunch_scripts/" . $Job->{"script"}; } @@ -855,37 +853,16 @@ else { eval { open(my $orig_manifest, '-|', 'arv-get', $collated_output) or die "failed to get collated manifest: $!"; - # Read the original manifest, and strip permission hints from it, - # so we can put the result in a Collection. - my @stripped_manifest_lines = (); my $orig_manifest_text = ''; while (my $manifest_line = <$orig_manifest>) { $orig_manifest_text .= $manifest_line; - my @words = split(/ /, $manifest_line, -1); - foreach my $ii (0..$#words) { - if ($words[$ii] =~ /^[0-9a-f]{32}\+/) { - $words[$ii] =~ s/\+A[0-9a-f]{40}@[0-9a-f]{8}\b//; - } - } - push(@stripped_manifest_lines, join(" ", @words)); } - my $stripped_manifest_text = join("", @stripped_manifest_lines); my $output = $arv->{'collections'}->{'create'}->execute('collection' => { - 'uuid' => md5_hex($stripped_manifest_text), 'manifest_text' => $orig_manifest_text, }); - Log(undef, "output " . $output->{uuid}); - $Job->update_attributes('output' => $output->{uuid}) if $job_has_uuid; - if ($Job->{'output_is_persistent'}) { - $arv->{'links'}->{'create'}->execute('link' => { - 'tail_kind' => 'arvados#user', - 'tail_uuid' => $User->{'uuid'}, - 'head_kind' => 'arvados#collection', - 'head_uuid' => $Job->{'output'}, - 'link_class' => 'resources', - 'name' => 'wants', - }); - } + Log(undef, "output uuid " . $output->{uuid}); + Log(undef, "output hash " . $output->{portable_data_hash}); + $Job->update_attributes('output' => $output->{portable_data_hash}) if $job_has_uuid; }; if ($@) { Log (undef, "Failed to register output manifest: $@"); @@ -1359,7 +1336,7 @@ sub save_meta return if $justcheckpoint; # checkpointing is not relevant post-Warehouse.pm $local_logfile->flush; - my $cmd = "arv-put --filename ''\Q$keep_logfile\E " + my $cmd = "arv-put --portable-data-hash --filename ''\Q$keep_logfile\E " . quotemeta($local_logfile->filename); my $loglocator = `$cmd`; die "system $cmd failed: $?" if $?; @@ -1500,15 +1477,29 @@ __DATA__ # checkout-and-build use Fcntl ':flock'; +use File::Path qw( make_path ); my $destdir = $ENV{"CRUNCH_SRC"}; my $commit = $ENV{"CRUNCH_SRC_COMMIT"}; my $repo = $ENV{"CRUNCH_SRC_URL"}; +my $task_work = $ENV{"TASK_WORK"}; + +for my $dir ($destdir, $task_work) { + if ($dir) { + make_path $dir; + -e $dir or die "Failed to create temporary directory ($dir): $!"; + } +} open L, ">", "$destdir.lock" or die "$destdir.lock: $!"; flock L, LOCK_EX; if (readlink ("$destdir.commit") eq $commit && -d $destdir) { - exit 0; + if (@ARGV) { + exec(@ARGV); + die "Cannot exec `@ARGV`: $!"; + } else { + exit 0; + } } unlink "$destdir.commit"; @@ -1554,7 +1545,12 @@ if ($commit) { close L; -exit 0; +if (@ARGV) { + exec(@ARGV); + die "Cannot exec `@ARGV`: $!"; +} else { + exit 0; +} sub shell_or_die {