# If this job requires a Docker image, install that.
my ($docker_locator, $docker_stream, $docker_hash, $docker_limitmem, $dockeruserarg);
if ($docker_locator = $Job->{docker_image_locator}) {
+ Log (undef, "Install docker image $docker_locator");
($docker_stream, $docker_hash) = find_docker_image($docker_locator);
if (!$docker_hash)
{
croak("No Docker image hash found from locator $docker_locator");
}
+ Log (undef, "docker image hash is $docker_hash");
$docker_stream =~ s/^\.//;
my $docker_install_script = qq{
if ! $docker_bin images -q --no-trunc --all | grep -qxF \Q$docker_hash\E; then
check_refresh_wanted();
check_squeue();
update_progress_stats();
- select (undef, undef, undef, 0.1);
}
elsif (time - $progress_stats_updated >= 30 || $progress_is_dirty)
{
update_progress_stats();
}
+ if (!$gotsome) {
+ select (undef, undef, undef, 0.1);
+ }
$working_slot_count = scalar(grep { $_->{node}->{fail_count} == 0 &&
$_->{node}->{hold_count} < 4 } @slot);
if (($thisround_failed_multiple >= 8 && $thisround_succeeded == 0) ||
foreach my $job (keys %reader)
{
my $buf;
- while (0 < sysread ($reader{$job}, $buf, 8192))
+ if (0 < sysread ($reader{$job}, $buf, 65536))
{
print STDERR $buf if $ENV{CRUNCH_DEBUG};
$jobstep[$job]->{stderr_at} = time;
$jobstep[$job]->{stderr} .= $buf;
+
+ # Consume everything up to the last \n
preprocess_stderr ($job);
+
if (length ($jobstep[$job]->{stderr}) > 16384)
{
- substr ($jobstep[$job]->{stderr}, 0, 8192) = "";
+ # If we get a lot of stderr without a newline, chop off the
+ # front to avoid letting our buffer grow indefinitely.
+ substr ($jobstep[$job]->{stderr},
+ 0, length($jobstep[$job]->{stderr}) - 8192) = "";
}
$gotsome = 1;
}
# whoa.
$main::please_freeze = 1;
}
- elsif ($line =~ /srun: error: Node failure on/) {
+ elsif ($line =~ /srun: error: (Node failure on|Aborting, .*\bio error\b)/) {
my $job_slot_index = $jobstep[$job]->{slotindex};
$slot[$job_slot_index]->{node}->{fail_count}++;
$jobstep[$job]->{tempfail} = 1;