6429: Add for test child container requests are set priority=0 when parent

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

diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job
index e2a4e264c3697e916ca8c88f98cf3a88d80e0e9a..b2c14d5b713dbfaa19c17b5d324ea4c81c1df9b8 100755 (executable)
--- a/sdk/cli/bin/crunch-job
+++ b/sdk/cli/bin/crunch-job
@@ -820,6 +820,9 @@ update_progress_stats();
 THISROUND:
 for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
 {
+  # Don't create new tasks if we already know the job's final result.
+  last if defined($main::success);
+
   my $id = $jobstep_todo[$todo_ptr];
   my $Jobstep = $jobstep[$id];
   if ($Jobstep->{level} != $level)
@@ -883,6 +886,11 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
 
     my $stdbuf = " stdbuf --output=0 --error=0 ";
 
+    my $arv_file_cache = "";
+    if (defined($Job->{'runtime_constraints'}->{'keep_cache_mb_per_task'})) {
+      $arv_file_cache = "--file-cache=" . ($Job->{'runtime_constraints'}->{'keep_cache_mb_per_task'} * 1024 * 1024);
+    }
+
     my $command =
        "if [ -e $ENV{TASK_WORK} ]; then rm -rf $ENV{TASK_WORK}; fi; "
         ."mkdir -p $ENV{CRUNCH_TMP} $ENV{JOB_WORK} $ENV{TASK_WORK} $ENV{TASK_KEEPMOUNT} "
@@ -893,7 +901,7 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
         .q{&& SWAP=$(awk '($1 == "SwapTotal:"){print $2}' </proc/meminfo) }
         ."&& MEMLIMIT=\$(( (\$MEM * 95) / ($ENV{CRUNCH_NODE_SLOTS} * 100) )) "
         ."&& let SWAPLIMIT=\$MEMLIMIT+\$SWAP ";
-    $command .= "&& exec arv-mount --by-pdh --allow-other $ENV{TASK_KEEPMOUNT} --exec ";
+    $command .= "&& exec arv-mount --by-pdh --crunchstat-interval=10 --allow-other $arv_file_cache $ENV{TASK_KEEPMOUNT} --exec ";
     if ($docker_hash)
     {
       my $containername = "$Jobstep->{arvados_task}->{uuid}-$Jobstep->{failures}";
@@ -962,7 +970,10 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
         $command .= $stdbuf;
         $command .= "perl - \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E";
       } else {
-        $command .= "/bin/sh -c \'mkdir -p \"$ENV{JOB_WORK}\" \"$ENV{TASK_WORK}\" && " .
+        $command .= "/bin/sh -c \'python -c " .
+            '"from pkg_resources import get_distribution as get; print \"Using Arvados SDK version\", get(\"arvados-python-client\").version"' .
+            ">&2 2>/dev/null; " .
+            "mkdir -p \"$ENV{JOB_WORK}\" \"$ENV{TASK_WORK}\" && " .
             "if which stdbuf >/dev/null ; then " .
             "  exec $stdbuf \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E ;" .
             " else " .
@@ -1018,7 +1029,7 @@ for (my $todo_ptr = 0; $todo_ptr <= $#jobstep_todo; $todo_ptr ++)
         ||
         ($round_num_freeslots > @freeslot && $todo_ptr+1 > $#jobstep_todo))
   {
-    last THISROUND if $main::please_freeze || defined($main::success);
+    last THISROUND if $main::please_freeze;
     if ($main::please_info)
     {
       $main::please_info = 0;
@@ -1179,6 +1190,9 @@ sub reapchildren
 
   if (!defined $task_success) {
     # task did not indicate one way or the other --> fail
+    Log($jobstepid, sprintf(
+          "ERROR: Task process exited %d, but never updated its task record to indicate success and record its output.",
+          exit_status_s($childstatus)));
     $Jobstep->{'arvados_task'}->{success} = 0;
     $Jobstep->{'arvados_task'}->save;
     $task_success = 0;
@@ -1695,20 +1709,24 @@ sub log_writer_finish()
 
   close($log_pipe_in);
 
+  my $logger_failed = 0;
   my $read_result = log_writer_read_output(120);
   if ($read_result == -1) {
+    $logger_failed = -1;
     Log (undef, "timed out reading from 'arv-put'");
   } elsif ($read_result != 0) {
+    $logger_failed = -2;
     Log(undef, "failed to read arv-put log manifest to EOF");
   }
 
   waitpid($log_pipe_pid, 0);
   if ($?) {
+    $logger_failed ||= $?;
     Log(undef, "log_writer_finish: arv-put exited " . exit_status_s($?))
   }
 
   close($log_pipe_out);
-  my $arv_put_output = $log_pipe_out_buf;
+  my $arv_put_output = $logger_failed ? undef : $log_pipe_out_buf;
   $log_pipe_pid = $log_pipe_in = $log_pipe_out = $log_pipe_out_buf =
       $log_pipe_out_select = undef;
 
@@ -1774,13 +1792,13 @@ sub save_meta
   my $justcheckpoint = shift; # false if this will be the last meta saved
   return if $justcheckpoint;  # checkpointing is not relevant post-Warehouse.pm
   return unless log_writer_is_active();
+  my $log_manifest = log_writer_finish();
+  return unless defined($log_manifest);
 
-  my $log_manifest = "";
   if ($Job->{log}) {
     my $prev_log_coll = api_call("collections/get", uuid => $Job->{log});
-    $log_manifest .= $prev_log_coll->{manifest_text};
+    $log_manifest = $prev_log_coll->{manifest_text} . $log_manifest;
   }
-  $log_manifest .= log_writer_finish();
 
   my $log_coll = api_call(
     "collections/create", ensure_unique_name => 1, collection => {
@@ -2159,10 +2177,11 @@ if (@ARGV) {
     $Log->("Built Python SDK virtualenv");
   }
 
-  my $pip_bin = "pip";
+  my @pysdk_version_cmd = ("python", "-c",
+    "from pkg_resources import get_distribution as get; print get('arvados-python-client').version");
   if ($venv_built) {
     $Log->("Running in Python SDK virtualenv");
-    $pip_bin = "$venv_dir/bin/pip";
+    @pysdk_version_cmd = ();
     my $orig_argv = join(" ", map { quotemeta($_); } @ARGV);
     @ARGV = ("/bin/sh", "-ec",
              ". \Q$venv_dir/bin/activate\E; exec $orig_argv");
@@ -2171,14 +2190,18 @@ if (@ARGV) {
            "\$PATH. Can't install Python SDK.");
   }
 
-  my $pkgs = `(\Q$pip_bin\E freeze 2>/dev/null | grep arvados) || dpkg-query --show '*arvados*'`;
-  if ($pkgs) {
-    $Log->("Using Arvados SDK:");
-    foreach my $line (split /\n/, $pkgs) {
-      $Log->($line);
+  if (@pysdk_version_cmd) {
+    open(my $pysdk_version_pipe, "-|", @pysdk_version_cmd);
+    my $pysdk_version = <$pysdk_version_pipe>;
+    close($pysdk_version_pipe);
+    if ($? == 0) {
+      chomp($pysdk_version);
+      $Log->("Using Arvados SDK version $pysdk_version");
+    } else {
+      # A lot could've gone wrong here, but pretty much all of it means that
+      # Python won't be able to load the Arvados SDK.
+      $Log->("Warning: Arvados SDK not found");
     }
-  } else {
-    $Log->("Arvados SDK packages not found");
   }
 
   while (my ($sdk_dir, $sdk_envkey) = each(%SDK_ENVVARS)) {