Expanded search path for cgroup stats, changed command line interface a bit.

[arvados.git] / sdk / cli / bin / crunch-job

diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job
index f092558cd75c9241c51625b4246a01ec8ed8dce0..5da8c78dda143cfac7befabd9fd6ba2610f5d21a 100755 (executable)
--- a/sdk/cli/bin/crunch-job
+++ b/sdk/cli/bin/crunch-job
@@ -76,6 +76,7 @@ use strict;
 use POSIX ':sys_wait_h';
 use Fcntl qw(F_GETFL F_SETFL O_NONBLOCK);
 use Arvados;
+use Digest::MD5 qw(md5_hex);
 use Getopt::Long;
 use IPC::Open2;
 use IO::Select;
@@ -640,7 +641,8 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
     $command .= "&& exec arv-mount --allow-other $ENV{TASK_KEEPMOUNT} --exec ";
     if ($docker_image)
     {
-      $command .= "$docker_bin run -i -a stdin -a stdout -a stderr ";
+      $command .= "crunchstat -cgroup-root=/sys/fs/cgroup -cgroup-parent=docker -cgroup-cid=$ENV{TASK_WORK}/docker.cid -poll=10000 ";
+      $command .= "$docker_bin run -i -a stdin -a stdout -a stderr --cidfile=$ENV{TASK_WORK}/docker.cid ";
       # Dynamically configure the container to use the host system as its
       # DNS server.  Get the host's global addresses from the ip command,
       # and turn them into docker --dns options using gawk.
@@ -653,10 +655,15 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
       }
       while (my ($env_key, $env_val) = each %ENV)
       {
-        $command .= "-e \Q$env_key=$env_val\E ";
+        if ($env_key =~ /^(JOB|TASK)_/) {
+          $command .= "-e \Q$env_key=$env_val\E ";
+        }
       }
       $command .= "\Q$docker_image\E ";
+    } else {
+      $command .= "crunchstat -cgroup-root=/sys/fs/cgroup -poll=10000 "
     }
+    $command .= "stdbuf -o0 -e0 ";
     $command .= "$ENV{CRUNCH_SRC}/crunch_scripts/" . $Job->{"script"};
     my @execargs = ('bash', '-c', $command);
     srun (\@srunargs, \@execargs, undef, $build_script_to_send);
@@ -799,21 +806,37 @@ goto ONELEVEL if !defined $main::success;
 
 release_allocation();
 freeze();
+my $collated_output = &collate_output();
+
 if ($job_has_uuid) {
-  $Job->update_attributes('output' => &collate_output(),
-                          'running' => 0,
-                          'success' => $Job->{'output'} && $main::success,
+  $Job->update_attributes('running' => 0,
+                          'success' => $collated_output && $main::success,
                           'finished_at' => scalar gmtime)
 }
 
-if ($Job->{'output'})
+if ($collated_output)
 {
   eval {
-    my $manifest_text = `arv keep get ''\Q$Job->{'output'}\E`;
-    $arv->{'collections'}->{'create'}->execute('collection' => {
-      'uuid' => $Job->{'output'},
+    open(my $orig_manifest, '-|', 'arv', 'keep', '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 @manifest_lines = ();
+    while (my $manifest_line = <$orig_manifest>) {
+      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(@manifest_lines, join(" ", @words));
+    }
+    my $manifest_text = join("", @manifest_lines);
+    my $output = $arv->{'collections'}->{'create'}->execute('collection' => {
+      'uuid' => md5_hex($manifest_text),
       'manifest_text' => $manifest_text,
     });
+    $Job->update_attributes('output' => $output->{uuid});
     if ($Job->{'output_is_persistent'}) {
       $arv->{'links'}->{'create'}->execute('link' => {
         'tail_kind' => 'arvados#user',