8782: Remove WIFEXITED check from crunch-job reapchildren.

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

diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job
index 3425519ba6f7cb4bbe56bc8373c7bd7b54d25120..86e018cc993bc7a29f16beef91a9f6c13bec4750 100755 (executable)
--- a/sdk/cli/bin/crunch-job
+++ b/sdk/cli/bin/crunch-job
@@ -432,7 +432,7 @@ fi
 
   # Determine whether this version of Docker supports memory+swap limits.
   ($exited, $stdout, $stderr) = srun_sync(
-    ["srun", "--nodelist=" . $node[0]],
+    ["srun", "--nodes=1"],
     [$docker_bin, 'run', '--help'],
     {label => "check --memory-swap feature"});
   $docker_limitmem = ($stdout =~ /--memory-swap/);
@@ -455,7 +455,7 @@ fi
       $try_user_arg = "--user=$try_user";
     }
     my ($exited, $stdout, $stderr) = srun_sync(
-      ["srun", "--nodelist=" . $node[0]],
+      ["srun", "--nodes=1"],
       ["/bin/sh", "-ec",
        "$docker_bin run $docker_run_args $try_user_arg $docker_hash id --user"],
       {label => $label});
@@ -852,7 +852,10 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
         .q{&& MEM=$(awk '($1 == "MemTotal:"){print $2}' </proc/meminfo) }
         .q{&& SWAP=$(awk '($1 == "SwapTotal:"){print $2}' </proc/meminfo) }
         ."&& MEMLIMIT=\$(( (\$MEM * 95) / ($ENV{CRUNCH_NODE_SLOTS} * 100) )) "
-        ."&& let SWAPLIMIT=\$MEMLIMIT+\$SWAP ";
+        ."&& let SWAPLIMIT=\$MEMLIMIT+\$SWAP "
+        ."&& if which crunchrunner >/dev/null ; then VOLUME_CRUNCHRUNNER=\"--volume=\$(which crunchrunner):/usr/local/bin/crunchrunner\" ; fi "
+        ."&& if test -f /etc/ssl/certs/ca-certificates.crt ; then VOLUME_CERTS=\"--volume=/etc/ssl/certs/ca-certificates.crt:/etc/arvados/ca-certificates.crt\" ; fi "
+        ."&& if test -f /etc/pki/tls/certs/ca-bundle.crt ; then VOLUME_CERTS=\"--volume=/etc/pki/tls/certs/ca-bundle.crt:/etc/arvados/ca-certificates.crt\" ; fi ";
 
     $command .= "&& exec arv-mount --read-write --mount-by-pdh=by_pdh --mount-tmp=tmp --crunchstat-interval=10 --allow-other $arv_file_cache \Q$keep_mnt\E --exec ";
     $ENV{TASK_KEEPMOUNT} = "$keep_mnt/by_pdh";
@@ -917,6 +920,10 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
       # For now, use the same approach as TASK_WORK above.
       $ENV{"JOB_WORK"} = "/tmp/crunch-job-work";
 
+      # Bind mount the crunchrunner binary and host TLS certificates file into
+      # the container.
+      $command .= "\"\$VOLUME_CRUNCHRUNNER\" \"\$VOLUME_CERTS\" ";
+
       while (my ($env_key, $env_val) = each %ENV)
       {
         if ($env_key =~ /^(ARVADOS|CRUNCH|JOB|TASK)_/) {
@@ -1145,13 +1152,6 @@ sub reapchildren
                     . $slot[$proc{$pid}->{slot}]->{cpu});
     my $jobstepidx = $proc{$pid}->{jobstepidx};
 
-    if (!WIFEXITED($childstatus))
-    {
-      # child did not exit (may be temporarily stopped)
-      Log ($jobstepidx, "child $pid did not actually exit in reapchildren, ignoring for now.");
-      next;
-    }
-
     $children_reaped++;
     my $elapsed = time - $proc{$pid}->{time};
     my $Jobstep = $jobstep[$jobstepidx];
@@ -1237,16 +1237,16 @@ sub reapchildren
     $progress_is_dirty = 1;
   }
 
-  if ($children_reaped > 0 && (scalar(@successful_task_uuids) == $children_reaped))
+  if (scalar(@successful_task_uuids) > 0)
   {
-    Log (undef, sprintf("Reaped %d children and all were successful, checking for new tasks from API server.", $children_reaped));
+    Log (undef, sprintf("%d tasks exited (%d succeeded), checking for new tasks from API server.", $children_reaped, scalar(@successful_task_uuids)));
     # Load new tasks
     my $newtask_list = [];
     my $newtask_results;
     do {
       $newtask_results = api_call(
         "job_tasks/list",
-        'filters' => '[["created_by_job_task_uuid","in",['.join(',',map {"\"$_\""} @successful_task_uuids).']]]',
+        'filters' => [["created_by_job_task_uuid","in",\@successful_task_uuids]],
         'order' => 'qsequence',
         'offset' => scalar(@$newtask_list),
           );