X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/0d62edcb9d25bf4dcdb20d8872ea7b438e12fc59..3b3c3a0869c2cf528b9e1c45c969bbbd47f6446e:/sdk/cli/bin/crunch-job

diff --git a/sdk/cli/bin/crunch-job b/sdk/cli/bin/crunch-job
index 40c9cf325c..35a3b1f23b 100755
--- a/sdk/cli/bin/crunch-job
+++ b/sdk/cli/bin/crunch-job
@@ -393,12 +393,12 @@ if (!defined $no_clear_tmp) {
   # Find FUSE mounts under $CRUNCH_TMP and unmount them.  Then clean
   # up work directories crunch_tmp/work, crunch_tmp/opt,
   # crunch_tmp/src*.
-  #
-  # TODO: When #5036 is done and widely deployed, we can limit mount's
-  # -t option to simply fuse.keep.
   my ($exited, $stdout, $stderr) = srun_sync(
     ["srun", "--nodelist=$nodelist", "-D", $ENV{'TMPDIR'}],
-    ['bash', '-ec', '-o', 'pipefail', 'mount -t fuse,fuse.keep | awk "(index(\$3, \"$CRUNCH_TMP\") == 1){print \$3}" | xargs -r -n 1 fusermount -u -z; sleep 1; rm -rf $JOB_WORK $CRUNCH_INSTALL $CRUNCH_TMP/task $CRUNCH_TMP/src* $CRUNCH_TMP/*.cid'],
+    ['bash', '-ec', q{
+arv-mount --unmount-timeout 10 --unmount-all ${CRUNCH_TMP}
+rm -rf ${JOB_WORK} ${CRUNCH_INSTALL} ${CRUNCH_TMP}/task ${CRUNCH_TMP}/src* ${CRUNCH_TMP}/*.cid
+    }],
     {label => "clean work dirs"});
   if ($exited != 0) {
     exit(EX_RETRY_UNLOCKED);
@@ -417,18 +417,22 @@ if ($docker_locator = $Job->{docker_image_locator}) {
   Log (undef, "docker image hash is $docker_hash");
   $docker_stream =~ s/^\.//;
   my $docker_install_script = qq{
-if $docker_bin images -q --no-trunc --all | grep -xF \Q$docker_hash\E >/dev/null; then
-    exit 0
+loaded() {
+  id=\$($docker_bin inspect --format="{{.ID}}" \Q$docker_hash\E) || return 1
+  echo "image ID is \$id"
+  [[ \${id} = \Q$docker_hash\E ]]
+}
+if loaded >&2 2>/dev/null; then
+  echo >&2 "image is already present"
+  exit 0
 fi
-declare -a exit_codes=("\${PIPESTATUS[@]}")
-if [ 0 != "\${exit_codes[0]}" ]; then
-   exit "\${exit_codes[0]}"  # `docker images` failed
-elif [ 1 != "\${exit_codes[1]}" ]; then
-   exit "\${exit_codes[1]}"  # `grep` encountered an error
-else
-   # Everything worked fine, but grep didn't find the image on this host.
-   arv-get \Q$docker_locator$docker_stream/$docker_hash.tar\E | $docker_bin load
+echo >&2 "docker image is not present; loading"
+arv-get \Q$docker_locator$docker_stream/$docker_hash.tar\E | $docker_bin load
+if ! loaded >&2; then
+  echo >&2 "`docker load` exited 0, but image is not found (!)"
+  exit 1
 fi
+echo >&2 "image loaded successfully"
 };
 
   my ($exited, $stdout, $stderr) = srun_sync(
@@ -715,6 +719,15 @@ foreach (split (/\n/, $Job->{knobs}))
 {
   Log (undef, "knob " . $_);
 }
+my $resp = api_call(
+  'nodes/list',
+  'filters' => [['hostname', 'in', \@node]],
+  'order' => 'hostname',
+  'limit' => scalar(@node),
+    );
+for my $n (@{$resp->{items}}) {
+  Log(undef, "$n->{hostname} $n->{uuid} ".JSON::encode_json($n->{properties}));
+}
 
 
 
@@ -1117,10 +1130,10 @@ freeze();
 my $collated_output = save_output_collection();
 Log (undef, "finish");
 
-save_meta();
+my $final_log = save_meta();
 
 my $final_state;
-if ($collated_output && $main::success) {
+if ($collated_output && $final_log && $main::success) {
   $final_state = 'Complete';
 } else {
   $final_state = 'Failed';
@@ -1593,9 +1606,10 @@ sub create_output_collection
 import arvados
 import sys
 print (arvados.api("v1").collections().
-       create(body={"manifest_text": sys.stdin.read()}).
+       create(body={"manifest_text": sys.stdin.read(),
+                    "owner_uuid": sys.argv[2]}).
        execute(num_retries=int(sys.argv[1]))["portable_data_hash"])
-}, retry_count());
+}, retry_count(), $Job->{owner_uuid});
 
   my $task_idx = -1;
   my $manifest_size = 0;
@@ -1746,7 +1760,7 @@ sub log_writer_start($)
   $log_pipe_pid = open2($log_pipe_out, $log_pipe_in,
                         'arv-put',
                         '--stream',
-                        '--retries', '3',
+                        '--retries', '6',
                         '--filename', $logfilename,
                         '-');
   $log_pipe_out_buf = "";
@@ -1780,7 +1794,7 @@ sub log_writer_finish()
   close($log_pipe_in);
 
   my $logger_failed = 0;
-  my $read_result = log_writer_read_output(120);
+  my $read_result = log_writer_read_output(600);
   if ($read_result == -1) {
     $logger_failed = -1;
     Log (undef, "timed out reading from 'arv-put'");
@@ -1843,6 +1857,7 @@ sub croak
   my ($package, $file, $line) = caller;
   my $message = "@_ at $file line $line\n";
   Log (undef, $message);
+  release_allocation();
   freeze() if @jobstep_todo;
   create_output_collection() if @jobstep_todo;
   cleanup();
@@ -1883,6 +1898,8 @@ sub save_meta
     });
   Log(undef, "log collection is " . $log_coll->{portable_data_hash});
   $Job->update_attributes('log' => $log_coll->{portable_data_hash});
+
+  return $log_coll->{portable_data_hash};
 }