-#!/usr/bin/perl
+#!/usr/bin/env perl
# -*- mode: perl; perl-indent-level: 2; indent-tabs-mode: nil; -*-
=head1 NAME
my $job_api_token;
my $no_clear_tmp;
my $resume_stash;
-my $docker_bin = "/usr/bin/docker.io";
+my $docker_bin = "docker.io";
GetOptions('force-unlock' => \$force_unlock,
'git-dir=s' => \$git_dir,
'job=s' => \$jobspec,
}
else
{
- $Job = JSON::decode_json($jobspec);
- $local_job = 1;
+ $local_job = JSON::decode_json($jobspec);
}
# at least able to run basic commands: they aren't down or severely
# misconfigured.
my $cmd = ['true'];
-if ($Job->{docker_image_locator}) {
+if (($Job || $local_job)->{docker_image_locator}) {
$cmd = [$docker_bin, 'ps', '-q'];
}
Log(undef, "Sanity check is `@$cmd`");
{
if (!$resume_stash)
{
- map { croak ("No $_ specified") unless $Job->{$_} }
+ map { croak ("No $_ specified") unless $local_job->{$_} }
qw(script script_version script_parameters);
}
- $Job->{'is_locked_by_uuid'} = $User->{'uuid'};
- $Job->{'started_at'} = gmtime;
- $Job->{'state'} = 'Running';
+ $local_job->{'is_locked_by_uuid'} = $User->{'uuid'};
+ $local_job->{'started_at'} = gmtime;
+ $local_job->{'state'} = 'Running';
- $Job = api_call("jobs/create", job => $Job);
+ $Job = api_call("jobs/create", job => $local_job);
}
$job_id = $Job->{'uuid'};
my $cleanpid = fork();
if ($cleanpid == 0)
{
- # Find FUSE mounts that look like Keep mounts (the mount path has the
- # word "keep") and unmount them. Then clean up work directories.
- # TODO: When #5036 is done and widely deployed, we can get rid of the
- # regular expression and just unmount everything with type fuse.keep.
+ # Find FUSE mounts under $CRUNCH_TMP and unmount them.
+ # Then clean up work directories.
+ # TODO: When #5036 is done and widely deployed, we can limit mount's
+ # -t option to simply fuse.keep.
srun (["srun", "--nodelist=$nodelist", "-D", $ENV{'TMPDIR'}],
- ['bash', '-ec', 'mount -t fuse,fuse.keep | awk \'($3 ~ /\ykeep\y/){print $3}\' | xargs -r -n 1 fusermount -u -z; sleep 1; rm -rf $JOB_WORK $CRUNCH_INSTALL $CRUNCH_TMP/task $CRUNCH_TMP/src* $CRUNCH_TMP/*.cid']);
+ ['bash', '-ec', '-o', 'pipefail', 'mount -t fuse,fuse.keep | awk "(index(\$3, \"$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* $CRUNCH_TMP/*.cid']);
exit (1);
}
while (1)
freeze_if_want_freeze ($cleanpid);
select (undef, undef, undef, 0.1);
}
- Log (undef, "Cleanup command exited ".exit_status_s($?));
+ if ($?) {
+ Log(undef, "Clean work dirs: exit ".exit_status_s($?));
+ exit(EX_RETRY_UNLOCKED);
+ }
}
# If this job requires a Docker image, install that.
-my ($docker_locator, $docker_stream, $docker_hash, $docker_limitmem);
+my ($docker_locator, $docker_stream, $docker_hash, $docker_limitmem, $dockeruserarg);
if ($docker_locator = $Job->{docker_image_locator}) {
($docker_stream, $docker_hash) = find_docker_image($docker_locator);
if (!$docker_hash)
{fork => 1});
$docker_limitmem = ($? == 0);
+ # Find a non-root Docker user to use.
+ # Tries the default user for the container, then 'crunch', then 'nobody',
+ # testing for whether the actual user id is non-zero. This defends against
+ # mistakes but not malice, but we intend to harden the security in the future
+ # so we don't want anyone getting used to their jobs running as root in their
+ # Docker containers.
+ my @tryusers = ("", "crunch", "nobody");
+ foreach my $try_user (@tryusers) {
+ my $try_user_arg;
+ if ($try_user eq "") {
+ Log(undef, "Checking if container default user is not UID 0");
+ $try_user_arg = "";
+ } else {
+ Log(undef, "Checking if user '$try_user' is not UID 0");
+ $try_user_arg = "--user=$try_user";
+ }
+ srun(["srun", "--nodelist=" . $node[0]],
+ ["/bin/sh", "-ec",
+ "a=`$docker_bin run --rm $try_user_arg $docker_hash id --user` && " .
+ " test \$a -ne 0"],
+ {fork => 1});
+ if ($? == 0) {
+ $dockeruserarg = $try_user_arg;
+ if ($try_user eq "") {
+ Log(undef, "Container will run with default user");
+ } else {
+ Log(undef, "Container will run with $dockeruserarg");
+ }
+ last;
+ }
+ }
+
+ if (!defined $dockeruserarg) {
+ croak("Could not find a user in container that is not UID 0 (tried default user, @tryusers) or there was a problem running 'id' in the container.");
+ }
+
if ($Job->{arvados_sdk_version}) {
# The job also specifies an Arvados SDK version. Add the SDKs to the
# tar file for the build script to install.
unless ($? == 0 && $sha1 =~ /^([0-9a-f]{40})$/) {
croak("`$gitcmd rev-list` exited "
.exit_status_s($?)
- .", '$treeish' not found. Giving up.");
+ .", '$treeish' not found, giving up");
}
$commit = $1;
Log(undef, "Version $treeish is commit $commit");
{
my $cidfile = "$ENV{CRUNCH_TMP}/$Jobstep->{arvados_task}->{uuid}-$Jobstep->{failures}.cid";
$command .= "crunchstat -cgroup-root=/sys/fs/cgroup -cgroup-parent=docker -cgroup-cid=$cidfile -poll=10000 ";
- $command .= "$docker_bin run --rm=true --attach=stdout --attach=stderr --attach=stdin -i --user=crunch --cidfile=$cidfile --sig-proxy ";
+ $command .= "$docker_bin run --rm=true --attach=stdout --attach=stderr --attach=stdin -i \Q$dockeruserarg\E --cidfile=$cidfile --sig-proxy ";
# We only set memory limits if Docker lets us limit both memory and swap.
# Memory limits alone have been supported longer, but subprocesses tend
# to get SIGKILL if they exceed that without any swap limit set.
$command .= "--memory=\${MEMLIMIT}k --memory-swap=\${SWAPLIMIT}k ";
}
- # Dynamically configure the container to use the host system as its
- # DNS server. Get the host's global addresses from the ip command,
- # and turn them into docker --dns options using gawk.
- $command .=
- q{$(ip -o address show scope global |
- gawk 'match($4, /^([0-9\.:]+)\//, x){print "--dns", x[1]}') };
-
# The source tree and $destdir directory (which we have
# installed on the worker host) are available in the container,
# under the same path.
}
$command .= "--env=\QHOME=$ENV{HOME}\E ";
$command .= "\Q$docker_hash\E ";
- $command .= "stdbuf --output=0 --error=0 ";
- $command .= "perl - $ENV{CRUNCH_SRC}/crunch_scripts/" . $Job->{"script"};
+
+ if ($Job->{arvados_sdk_version}) {
+ $command .= "stdbuf --output=0 --error=0 ";
+ $command .= "perl - \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E";
+ } else {
+ $command .= "/bin/sh -c \'mkdir -p \"$ENV{JOB_WORK}\" \"$ENV{TASK_WORK}\" && exec \Q$ENV{CRUNCH_SRC}/crunch_scripts/$Job->{script}\E\'";
+ }
} else {
# Non-docker run
$command .= "crunchstat -cgroup-root=/sys/fs/cgroup -poll=10000 ";
my $gotsome
= readfrompipes ()
+ reapchildren ();
- if (!$gotsome)
+ if (!$gotsome || ($latest_refresh + 2 < scalar time))
{
check_refresh_wanted();
check_squeue();
update_progress_stats();
select (undef, undef, undef, 0.1);
}
- elsif (time - $progress_stats_updated >= 30)
+ elsif (time - $progress_stats_updated >= 30 || $progress_is_dirty)
{
update_progress_stats();
}
$progress_stats_updated = time;
return if !$progress_is_dirty;
my ($todo, $done, $running) = (scalar @jobstep_todo,
- scalar @jobstep_done,
- scalar @slot - scalar @freeslot - scalar @holdslot);
+ scalar @jobstep_done,
+ scalar keys(%proc));
$Job->{'tasks_summary'} ||= {};
$Job->{'tasks_summary'}->{'todo'} = $todo;
$Job->{'tasks_summary'}->{'done'} = $done;
}
__DATA__
-#!/usr/bin/perl
+#!/usr/bin/env perl
#
# This is crunch-job's internal dispatch script. crunch-job running on the API
# server invokes this script on individual compute nodes, or localhost if we're
my $python_dir = "$install_dir/python";
if ((-d $python_dir) and can_run("python2.7")) {
open(my $egg_info_pipe, "-|",
- "python2.7 \Q$python_dir/setup.py\E --quiet egg_info 2>&1 >/dev/null");
+ "python2.7 \Q$python_dir/setup.py\E egg_info 2>&1 >/dev/null");
my @egg_info_errors = <$egg_info_pipe>;
close($egg_info_pipe);
+
if ($?) {
- if (@egg_info_errors and ($egg_info_errors[-1] =~ /\bgit\b/)) {
+ if (@egg_info_errors and (($egg_info_errors[-1] =~ /\bgit\b/) or ($egg_info_errors[-1] =~ /\[Errno 2\]/))) {
# egg_info apparently failed because it couldn't ask git for a build tag.
# Specify no build tag.
open(my $pysdk_cfg, ">>", "$python_dir/setup.cfg");
} else {
my $egg_info_exit = $? >> 8;
foreach my $errline (@egg_info_errors) {
- print STDERR_ORIG $errline;
+ warn $errline;
}
warn "python setup.py egg_info failed: exit $egg_info_exit";
exit ($egg_info_exit || 1);
projects / arvados.git / blobdiff