Skip to content

Commit

Permalink
Merge branch 'frontend_machinectl' into '2.5'
Browse files Browse the repository at this point in the history 
[oarsub] Use systemd-run when running processes on the frontend for deploy/cosystem jobs (bug 13345)

See merge request grid5000/oar!28
  • Loading branch information
Pierre Neyron committed Nov 25, 2021
2 parents 7401a1c + 0332e16 commit 662a6b0
Show file tree
Hide file tree
Showing 4 changed files with 51 additions and 4 deletions.
10 changes: 9 additions & 1 deletion sources/core/modules/runner/bipbip.in
View file
Original file line number Diff line number Diff line change
Expand Up @@ -382,11 +382,18 @@ my $epilogue_exec_file;
if (is_conf("EPILOGUE_EXEC_FILE")){
$epilogue_exec_file = get_conf("EPILOGUE_EXEC_FILE");
}

my $passive_job_hook_exec_file;
if (is_conf("PASSIVE_JOB_HOOK_EXEC_FILE")){
$passive_job_hook_exec_file = get_conf("PASSIVE_JOB_HOOK_EXEC_FILE");
}
my $deploy_cosystem_job_exec_system;
if (is_conf("DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM")){
$deploy_cosystem_job_exec_system = get_conf("DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM");
if (($deploy_cosystem_job_exec_system ne "none") and ($deploy_cosystem_job_exec_system ne "systemd-run")) {
oar_error("[bipbip $Job_id] Invalid configuration for DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM: '$deploy_cosystem_job_exec_system' is unsupported\n");
# Don't exit, because it causes the job to be stuck in 'Launching' state...
}
}
my @oarexec_files = ($OAR_Tools_module_filepath,"$ENV{OARDIR}/oarexec");

my $host_to_connect_via_ssh = $hosts[0];
Expand Down Expand Up @@ -436,6 +443,7 @@ my %data_to_transfer = (
prologue => $prologue_exec_file,
epilogue => $epilogue_exec_file,
passive_job_hook => $passive_job_hook_exec_file,
deploy_cosystem_job_exec_system => $deploy_cosystem_job_exec_system,
tmp_directory => OAR::Tools::get_default_oarexec_directory,
detach_oarexec => $Detach_oarexec,
cpuset_full_path => $oarexec_cpuset_path
Expand Down
12 changes: 11 additions & 1 deletion sources/core/modules/runner/oarexec
View file
Original file line number Diff line number Diff line change
Expand Up @@ -298,13 +298,23 @@ if (!defined($shell)){
print("[oarexec $Job_id] error: user $Job->{job_user} does not exist on this node, $host\n");
exit(7);
}
my $uid = $pass_info[2];

my @cmd;
my $stdin_script_to_send;
my $oardo_become_user = $Job->{job_user};
if ( $Job->{mode} eq "PASSIVE" ){
my $str;
($str, $stdin_script_to_send) = OAR::Tools::get_oarexecuser_perl_script_for_oarexec($Node_file,$Job_id,$Job->{array_id},$Job->{array_index},$Job->{user},$shell,$Job->{launching_directory},$Job->{stdout_file},$Job->{stderr_file},$Res_file,$Job->{name},$Job->{project},$Job->{walltime},$Job->{walltime_seconds},$Job->{job_env},$Job->{command});
@cmd = ("oardodo","perl","-e",$str);
if ((defined($Job->{types}->{deploy}) or defined($Job->{types}->{cosystem})) and defined($Job->{deploy_cosystem_job_exec_system})) {
if ($Job->{deploy_cosystem_job_exec_system} eq "systemd-run") {
$oardo_become_user = "";
@cmd = ("oardodo","systemd-run","-q","--uid=$uid","--pipe","--wait","--collect","--service-type=exec","--slice=user-$uid.slice","perl","-e",$str);
} elsif ($Job->{deploy_cosystem_job_exec_system} eq "none") {
# Use regular command by default
}
}
}

#resolve terminal type problems
Expand Down Expand Up @@ -463,7 +473,7 @@ if ($Job->{mode} eq "PASSIVE"){
close(STDIN);
open(STDIN, "<& pipe_stdin_read");

$ENV{OARDO_BECOME_USER} = $Job->{job_user};
$ENV{OARDO_BECOME_USER} = $oardo_become_user;
umask(oct($Old_umask));
exec(@cmd);
warn("[ERROR] Cannot find @cmd\n");
Expand Down
24 changes: 22 additions & 2 deletions sources/core/qfunctions/oarsub
View file
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,7 @@ my $remote_host;
my $remote_port;
my $Deploy_hostname;
my $Cosystem_hostname;
my $Deploy_cosystem_job_exec_system;
my $Cpuset_field;
my $Cpuset_path;
my $Host = hostname ;
Expand Down Expand Up @@ -165,6 +166,7 @@ sub connect_job($$$){
OAR::Sub::close_db_connection();
my @passinfo = getpwnam($lusr) or die("Cannot retreive system information for user $lusr\n");
my $shell=$passinfo[8];
my $uid=$passinfo[2];
unless ((defined($xauth_path)) and (-x $xauth_path) and ($ENV{DISPLAY} =~ /^[\w.-]*:\d+(?:\.\d+)?$/)) {
$ENV{DISPLAY}="";
}
Expand Down Expand Up @@ -222,10 +224,20 @@ sub connect_job($$$){
if (defined($interactive_job_hook_file)) {
$interactive_job_hook_cmd = "if test -x $interactive_job_hook_file; then oardodo $interactive_job_hook_file \$\$,\$PPID:$lusr:$job_id:$job->{name}:$job->{project}:$moldable->{moldable_walltime}; else true; fi && ";
}

my $oardodo_cmd = "OARDO_BECOME_USER=$job_user oardodo bash --noprofile --norc -c '$str'";
if (defined($types->{cosystem}) or defined($types->{deploy}) or ($#hosts < 0)) {
if ($Deploy_cosystem_job_exec_system eq "systemd-run") {
$oardodo_cmd = "oardodo systemd-run -q --uid=$uid --pty --wait --collect --service-type=exec --slice=user-$uid.slice /bin/bash --noprofile --norc -c '$str'";
} elsif ($Deploy_cosystem_job_exec_system eq "none") {
# Use regular command by default
}
}

if ($ENV{DISPLAY} ne ""){
$cmd[$i] = "bash -c '${interactive_job_hook_cmd}echo \$PPID >> $oarsub_pids && ($xauth_path -q extract - \${DISPLAY/#localhost:/:} | OARDO_BECOME_USER=$lusr oardodo $xauth_path merge -) && [ \"$lusr\" != \"$job_user\" ] && OARDO_BECOME_USER=$lusr oardodo bash --noprofile --norc -c \"chmod 660 \\\$HOME/.Xauthority\" ;TTY=\$(tty) && test -e \$TTY && oardodo chown $job_user:oar \$TTY && oardodo chmod 660 \$TTY' && OARDO_BECOME_USER=$job_user oardodo bash --noprofile --norc -c '$str'";$i++;
$cmd[$i] = "bash -c '${interactive_job_hook_cmd}echo \$PPID >> $oarsub_pids && ($xauth_path -q extract - \${DISPLAY/#localhost:/:} | OARDO_BECOME_USER=$lusr oardodo $xauth_path merge -) && [ \"$lusr\" != \"$job_user\" ] && OARDO_BECOME_USER=$lusr oardodo bash --noprofile --norc -c \"chmod 660 \\\$HOME/.Xauthority\" ;TTY=\$(tty) && test -e \$TTY && oardodo chown $job_user:oar \$TTY && oardodo chmod 660 \$TTY' && $oardodo_cmd";$i++;
}else{
$cmd[$i] = "bash -c '${interactive_job_hook_cmd}echo \$PPID >> $oarsub_pids && TTY=\$(tty) && test -e \$TTY && oardodo chown $job_user:oar \$TTY && oardodo chmod 660 \$TTY' && OARDO_BECOME_USER=$job_user oardodo bash --noprofile --norc -c '$str'";$i++;
$cmd[$i] = "bash -c '${interactive_job_hook_cmd}echo \$PPID >> $oarsub_pids && TTY=\$(tty) && test -e \$TTY && oardodo chown $job_user:oar \$TTY && oardodo chmod 660 \$TTY' && $oardodo_cmd";$i++;
}

#essential: you become oar instead of the user
Expand Down Expand Up @@ -553,6 +565,14 @@ if (!defined($Cosystem_hostname)){
$Cosystem_hostname = $remote_host;
}

$Deploy_cosystem_job_exec_system = get_conf("DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM");
if (!defined($Deploy_cosystem_job_exec_system)) {
$Deploy_cosystem_job_exec_system = "none";
}
if (($Deploy_cosystem_job_exec_system ne "none") and ($Deploy_cosystem_job_exec_system ne "systemd-run")) {
die("[ERROR] Bad configuration for DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM: '$Deploy_cosystem_job_exec_system' is unsupported\n");
}

$Cpuset_field = get_conf("JOB_RESOURCE_MANAGER_PROPERTY_DB_FIELD");
$Cpuset_path = get_conf("CPUSET_PATH");

Expand Down
9 changes: 9 additions & 0 deletions sources/core/tools/oar.conf.in
View file
Original file line number Diff line number Diff line change
Expand Up @@ -87,6 +87,15 @@ DEPLOY_HOSTNAME="127.0.0.1"
COSYSTEM_HOSTNAME="127.0.0.1"
# Configuration for module: server, user, api

# Specify what system to use for the execution on the deploy and cosystem frontends
# Possible values are:
# none: just run in a bash shell (default)
# systemd-run: use systemd-run
# systemd-run executes the job processes in user slices managed by systemd, which
# could easily allow to set resource limits through cgroups.
#DEPLOY_COSYSTEM_JOB_EXEC_SYSTEM="systemd-run"
# Configuration for module: server, user, api

# Specify the database field to use to fill the file on the first node of the
# job in $OAR_NODE_FILE (default is 'network_address'). Only resources with
# type=default are displayed in this file.
Expand Down

0 comments on commit 662a6b0

Please sign in to comment.