arvados-cli 0.1.20150224142231 → 0.1.20150227192315

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: f48e5cd2ae93dc785d14240c4d586269994f26c7
-  data.tar.gz: f49bf2da971cf11912f90dfca0c9efd91550d37c
+  metadata.gz: fee9adf2c5e05f6708bcac31dde6f59bf2fe0247
+  data.tar.gz: 32b1e5d39a1e99b697efdb7c4a46f071c241ef4c
 SHA512:
-  metadata.gz: 2ecc3650403a6e87f0c36b2bc629015841703bf24abe3092e6d51ba2551bca8084b365f18cc9f0bd559d0d9e8121875ed00a9c3dceba64af52343c9ac15e1ce2
-  data.tar.gz: 030f93979b6a408a9b87a91ced431ec0ea10876e0bdc7ed0570241017537dc0a6ea1447ade6e22ff24b50919a9bd86fa8df0b9655f137b1d088f9ecbb8120f5b
+  metadata.gz: a53ba2e71b329d22e97800f5aca9bd765f63e4155d17c4ca1883b0939dbc411a29591fa33fee741c04975b9ee4533a3b2ab0bd8814f237e29b549e32f49639e6
+  data.tar.gz: aab8e8575cb748acb045b39ad2d24ba0ef503468e3a503c3529fc98eb21ea44334dfc8a22eecc3615649af496b40142ad03661386c3cba4746ea8c7db5ff5966

data/bin/crunch-job

@@ -941,7 +941,7 @@ if (!$collated_output) {
   Log (undef, "Failed to write output collection");
 }
 else {
-  Log(undef, "output hash " . $collated_output);
+  Log(undef, "job output $collated_output");
   $Job->update_attributes('output' => $collated_output);
 }
 
@@ -1037,7 +1037,6 @@ sub reapchildren
     if (!$temporary_fail || $Jobstep->{'failures'} >= 3) {
       # Give up on this task, and the whole job
       $main::success = 0;
-      $main::please_freeze = 1;
     }
     # Put this task back on the todo queue
     push @jobstep_todo, $jobstepid;
@@ -1056,7 +1055,9 @@ sub reapchildren
   $Jobstep->{'arvados_task'}->{finished_at} = strftime "%Y-%m-%dT%H:%M:%SZ", gmtime($Jobstep->{finishtime});
   $Jobstep->{'arvados_task'}->save;
   process_stderr ($jobstepid, $task_success);
-  Log ($jobstepid, "output " . $Jobstep->{'arvados_task'}->{output});
+  Log ($jobstepid, sprintf("task output (%d bytes): %s",
+                           length($Jobstep->{'arvados_task'}->{output}),
+                           $Jobstep->{'arvados_task'}->{output}));
 
   close $reader{$jobstepid};
   delete $reader{$jobstepid};
@@ -1252,16 +1253,19 @@ sub process_stderr
 sub fetch_block
 {
   my $hash = shift;
-  my ($keep, $child_out, $output_block);
-
-  my $cmd = "arv-get \Q$hash\E";
-  open($keep, '-|', $cmd) or die "fetch_block: $cmd: $!";
-  $output_block = '';
+  my $keep;
+  if (!open($keep, "-|", "arv-get", "--retries", retry_count(), $hash)) {
+    Log(undef, "fetch_block run error from arv-get $hash: $!");
+    return undef;
+  }
+  my $output_block = "";
   while (1) {
     my $buf;
     my $bytes = sysread($keep, $buf, 1024 * 1024);
     if (!defined $bytes) {
-      die "reading from arv-get: $!";
+      Log(undef, "fetch_block read error from arv-get: $!");
+      $output_block = undef;
+      last;
     } elsif ($bytes == 0) {
       # sysread returns 0 at the end of the pipe.
       last;
@@ -1271,6 +1275,10 @@ sub fetch_block
     }
   }
   close $keep;
+  if ($?) {
+    Log(undef, "fetch_block arv-get exited " . exit_status_s($?));
+    $output_block = undef;
+  }
   return $output_block;
 }
 
@@ -1283,50 +1291,64 @@ sub create_output_collection
   Log (undef, "collate");
 
   my ($child_out, $child_in);
-  my $pid = open2($child_out, $child_in, 'python', '-c',
-                  'import arvados; ' .
-                  'import sys; ' .
-                  'print arvados.api()' .
-                  '.collections()' .
-                  '.create(body={"manifest_text":sys.stdin.read()})' .
-                  '.execute()["portable_data_hash"]'
-      );
+  my $pid = open2($child_out, $child_in, 'python', '-c', q{
+import arvados
+import sys
+print (arvados.api("v1").collections().
+       create(body={"manifest_text": sys.stdin.read()}).
+       execute(num_retries=int(sys.argv[1]))["portable_data_hash"])
+}, retry_count());
 
   my $task_idx = -1;
+  my $manifest_size = 0;
   for (@jobstep)
   {
     ++$task_idx;
-    next unless exists $_->{'arvados_task'}->{'output'};
     my $output = $_->{'arvados_task'}->{output};
-    if ($output !~ /^[0-9a-f]{32}(\+\S+)*$/)
-    {
-      print $child_in $output;
-    }
-    elsif (defined (my $outblock = fetch_block ($output)))
-    {
-      print $child_in $outblock;
+    next if (!defined($output));
+    my $next_write;
+    if ($output =~ /^[0-9a-f]{32}(\+\S+)*$/) {
+      $next_write = fetch_block($output);
+    } else {
+      $next_write = $output;
     }
-    else
-    {
+    if (defined($next_write)) {
+      if (!defined(syswrite($child_in, $next_write))) {
+        # There's been an error writing.  Stop the loop.
+        # We'll log details about the exit code later.
+        last;
+      } else {
+        $manifest_size += length($next_write);
+      }
+    } else {
       my $uuid = $_->{'arvados_task'}->{'uuid'};
       Log (undef, "Error retrieving '$output' output by task $task_idx ($uuid)");
       $main::success = 0;
     }
   }
-  $child_in->close;
+  close($child_in);
+  Log(undef, "collated output manifest text to send to API server is $manifest_size bytes with access tokens");
 
   my $joboutput;
   my $s = IO::Select->new($child_out);
   if ($s->can_read(120)) {
-    sysread($child_out, $joboutput, 64 * 1024 * 1024);
-    chomp($joboutput);
-    # TODO: Ensure exit status == 0.
+    sysread($child_out, $joboutput, 1024 * 1024);
+    waitpid($pid, 0);
+    if ($?) {
+      Log(undef, "output collection creation exited " . exit_status_s($?));
+      $joboutput = undef;
+    } else {
+      chomp($joboutput);
+    }
   } else {
     Log (undef, "timed out while creating output collection");
+    foreach my $signal (2, 2, 2, 15, 15, 9) {
+      kill($signal, $pid);
+      last if waitpid($pid, WNOHANG) == -1;
+      sleep(1);
+    }
   }
-  # TODO: kill $pid instead of waiting, now that we've decided to
-  # ignore further output.
-  waitpid($pid, 0);
+  close($child_out);
 
   return $joboutput;
 }

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: arvados-cli
 version: !ruby/object:Gem::Version
-  version: 0.1.20150224142231
+  version: 0.1.20150227192315
 platform: ruby
 authors:
 - Arvados Authors
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2015-02-24 00:00:00.000000000 Z
+date: 2015-02-27 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: arvados
@@ -178,7 +178,7 @@ dependencies:
     - - "<"
       - !ruby/object:Gem::Version
         version: 1.0.0
-description: Arvados command line tools, git commit 18ddae75beb91b0792ad34a9b4996316715e67b7
+description: Arvados command line tools, git commit 62fea2a463eb0d8c3aef84a56ba84561a196666b
 email: gem-dev@curoverse.com
 executables:
 - arv