[slurm-users] Slurmd enabled crash with CgroupV2

Fri Mar 10 15:41:24 UTC 2023

Hello to all,

I'm trying to do an installation of Slurm with cgroupv2 activated.

But I'm facing an odd thing : when slurmd is enabled it crash at the next reboot and will never start unless i disable it.

Here is a full example of the situation

[root at compute ~]# systemctl start slurmd
[root at compute ~]# systemctl status slurmd
● slurmd.service - Slurm node daemon
   Loaded: loaded (/usr/lib/systemd/system/slurmd.service; disabled; vendor preset: disabled)
   Active: active (running) since Fri 2023-03-10 15:57:00 CET; 967ms ago
 Main PID: 8053 (slurmd)
    Tasks: 1
   Memory: 3.1M
   CGroup: /system.slice/slurmd.service
           └─8053 /opt/slurm_bin/sbin/slurmd -D --conf-server XXXXX:6817 -s

mars 10 15:57:00 compute.cluster.lab systemd[1]: Started Slurm node daemon.
mars 10 15:57:00 compute.cluster.lab slurmd[8053]: slurmd: slurmd version 23.02.0 started
mars 10 15:57:00 compute.cluster.lab slurmd[8053]: slurmd: slurmd started on Fri, 10 Mar 2023 15:57:00 +0100
mars 10 15:57:00 compute.cluster.lab slurmd[8053]: slurmd: CPUs=48 Boards=1 Sockets=2 Cores=24 Threads=1 Memory=385311 TmpDisk=19990 Uptime=12>

[root at compute ~]# systemctl enable slurmd
Created symlink /etc/systemd/system/multi-user.target.wants/slurmd.service → /usr/lib/systemd/system/slurmd.service.

[root at compute ~]#  reboot now

> [ reboot of the node]

[adm at compute ~]$ sudo systemctl status slurmd
● slurmd.service - Slurm node daemon
   Loaded: loaded (/usr/lib/systemd/system/slurmd.service; enabled; vendor preset: disabled)
   Active: failed (Result: exit-code) since Fri 2023-03-10 16:00:33 CET; 1min 0s ago
  Process: 2659 ExecStart=/opt/slurm_bin/sbin/slurmd -D --conf-server XXXX:6817 -s $SLURMD_OPTIONS (code=exited, status=1/FAILURE)
 Main PID: 2659 (code=exited, status=1/FAILURE)
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: slurmd version 23.02.0 started
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: Controller cpuset is not enabled!
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: Controller cpu is not enabled!
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: cpu cgroup controller is not available.
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: There's an issue initializing memory or cpu controller
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: Couldn't load specified plugin name for jobacct_gather/cgroup: Plugin init()>
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: error: cannot create jobacct_gather context for jobacct_gather/cgroup
mars 10 16:00:33 compute.cluster.lab slurmd[2659]: slurmd: fatal: Unable to initialize jobacct_gather
mars 10 16:00:33 compute.cluster.lab systemd[1]: slurmd.service: Main process exited, code=exited, status=1/FAILURE
mars 10 16:00:33 compute.cluster.lab systemd[1]: slurmd.service: Failed with result 'exit-code'.

[adm at compute ~]$ sudo systemctl start slurmd
[adm at compute ~]$ sudo systemctl status slurmd
● slurmd.service - Slurm node daemon
   Loaded: loaded (/usr/lib/systemd/system/slurmd.service; enabled; vendor preset: disabled)
   Active: failed (Result: exit-code) since Fri 2023-03-10 16:01:37 CET; 1s ago
  Process: 3321 ExecStart=/opt/slurm_bin/sbin/slurmd -D --conf-server XXXX:6817 -s $SLURMD_OPTIONS (code=exited, status=1/FAILURE)
 Main PID: 3321 (code=exited, status=1/FAILURE)
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: slurmd version 23.02.0 started
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: Controller cpuset is not enabled!
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: Controller cpu is not enabled!
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: cpu cgroup controller is not available.
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: There's an issue initializing memory or cpu controller
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: Couldn't load specified plugin name for jobacct_gather/cgroup: Plugin init()>
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: error: cannot create jobacct_gather context for jobacct_gather/cgroup
mars 10 16:01:37 compute.cluster.lab slurmd[3321]: slurmd: fatal: Unable to initialize jobacct_gather
mars 10 16:01:37 compute.cluster.lab systemd[1]: slurmd.service: Main process exited, code=exited, status=1/FAILURE
mars 10 16:01:37 compute.cluster.lab systemd[1]: slurmd.service: Failed with result 'exit-code'.

[adm at compute ~]$ sudo systemctl disable slurmd
Removed /etc/systemd/system/multi-user.target.wants/slurmd.service.
[adm at compute ~]$ sudo systemctl start slurmd
[adm at compute ~]$ sudo systemctl status slurmd
● slurmd.service - Slurm node daemon
   Loaded: loaded (/usr/lib/systemd/system/slurmd.service; disabled; vendor preset: disabled)
   Active: active (running) since Fri 2023-03-10 16:01:45 CET; 1s ago
 Main PID: 3358 (slurmd)
    Tasks: 1
   Memory: 6.1M
   CGroup: /system.slice/slurmd.service
           └─3358 /opt/slurm_bin/sbin/slurmd -D --conf-server XXXX:6817 -s
mars 10 16:01:45 compute.cluster.lab systemd[1]: Started Slurm node daemon.
mars 10 16:01:45 compute.cluster.lab slurmd[3358]: slurmd: slurmd version 23.02.0 started
mars 10 16:01:45 compute.cluster.lab slurmd[3358]: slurmd: slurmd started on Fri, 10 Mar 2023 16:01:45 +0100
mars 10 16:01:45 compute.cluster.lab slurmd[3358]: slurmd: CPUs=48 Boards=1 Sockets=2 Cores=24 Threads=1 Memory=385311 TmpDisk=19990 Uptime=84>

As you can see. Slurmd successfully start only when not enable after a reboot.

- I'm using Rocky Linux 8  and I've configured cgroupv2 with grubby

> grubby --update-kernel=ALL --args="systemd.unified_cgroup_hierarchy=1 systemd.legacy_systemd_cgroup_controller=0 cgroup_no_v1=all"

- Slurm 23.02 is build with rpmbuild and slurmd on the compute node is installed with rpm

- Here is my cgroup.conf :

CgroupPlugin=cgroup/v2
ConstrainCores=yes
ConstrainRAMSpace=yes
ConstrainSwapSpace=yes
ConstrainDevices=no

And my slurm.conf have :

ProctrackType=proctrack/cgroup
TaskPlugin=task/cgroup,task/affinity
JobAcctGatherType=jobacct_gather/cgroup

- If i do "systemctl start slurmd" on a compute node it's a success.

- If i do "systemctl enable slurmd" and then "systemctl restart slurmd" it's still ok

- if i enable and reboot, slurmd send this error :

slurmd: error: Controller cpuset is not enabled!
slurmd: error: Controller cpu is not enabled!
slurmd: error: cpu cgroup controller is not available.
slurmd: error: There's an issue initializing memory or cpu controlle

-  I've done some research and read about cgroup.subtree_control. And so if i do:

cat /sys/fs/cgroup/cgroup.subtree_control
memory pids

So I've tried to follow the RedHat documentation with there example : ( the link of the RedHat page here <https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/managing_monitoring_and_updating_the_kernel/using-cgroups-v2-to-control-distribution-of-cpu-time-for-applications_managing-monitoring-and-updating-the-kernel> )

echo "+cpu" >> /sys/fs/cgroup/cgroup.subtree_control
echo "+cpuset" >> /sys/fs/cgroup/cgroup.subtree_control

cat /sys/fs/cgroup/cgroup.subtree_control
cpuset cpu memory pids

And indeed i can restart slurmd.

But at the next boot it failed again  and /sys/fs/cgroup/cgroup.subtree_control is back with "memory pids" only.

And strangely i found if slurmd is enabled and then i disable it, it change the value of /sys/fs/cgroup/cgroup.subtree_control :

[root at compute ~]# cat /sys/fs/cgroup/cgroup.subtree_control
memory pids
[root at compute ~]# systemctl disable slurmd
Removed /etc/systemd/system/multi-user.target.wants/slurmd.service.
[root at compute ~]# cat /sys/fs/cgroup/cgroup.subtree_control
cpuset cpu io memory pids

I've made a script at launch time as a dirty fix by using ExecPreStart in slurmd.service:

ExecStartPre=/opt/slurm_bin/dirty_fix_slurmd.sh

with dirty_fix_slurmd.sh:

#!/bin//bash
echo "+cpu" >> /sys/fs/cgroup/cgroup.subtree_control
echo "+cpuset" >> /sys/fs/cgroup/cgroup.subtree_control
echo "+cpu" >> /sys/fs/cgroup/system.slice/cgroup.subtree_control
echo "+cpuset" >> /sys/fs/cgroup/system.slice/cgroup.subtree_control

(And i'm not sure if this is something good to do ?)

If you have an idea how to correct this situation

Have a nice day

Thank you
Tristan LEFEBVRE
CONFIDENTIALITE : ce courriel et les éventuelles pièces attachées sont la propriété de l’IRT Jules Verne, sont confidentiels et sont réservés à l’usage de la ou des personne(s) identifées(s) comme destinataire(s). Si vous avez reçu ce courriel par erreur, toute utilisation, divulgation, ou copie de ce courriel est interdite. Dans ce cas, merci d’en informer immédiatement l'expéditeur et de supprimer le courriel et ses pièces jointes.
CONFIDENTIALITY : This e-mail and any attachments are IRT Jules Verne’s property and are intended solely for the person or entity to whom it is addressed, and may contain confidential or privileged information. Should you have received this e-mail in error, any use, disclosure, or copy of this email is prohibited. In this case, please inform the sender immediately and delete this email and its attachments.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20230310/3981b625/attachment-0001.htm>