Step 6 – (Optional) Test the connection

Confirm that the sackd service is running. Sample output follows. If there are errors, they will commonly show up here.

[root@ip-10-3-27-112 ~]# systemctl status sackd
[x] sackd.service - Slurm auth and cred kiosk daemon
   Loaded: loaded (/etc/systemd/system/sackd.service; enabled; vendor preset: disabled)
   Active: active (running) since Tue 2024-07-16 16:34:55 UTC; 8s ago
 Main PID: 9985 (sackd)
   CGroup: /system.slice/sackd.service
           └─9985 /opt/aws/pcs/scheduler/slurm-23.11/sbin/sackd --systemd --conf-server=10.3.149.220:6817

Jul 16 16:34:55 ip-10-3-27-112.ec2.internal systemd[1]: Starting Slurm auth and cred kiosk daemon...
Jul 16 16:34:55 ip-10-3-27-112.ec2.internal systemd[1]: Started Slurm auth and cred kiosk daemon.
Jul 16 16:34:55 ip-10-3-27-112.ec2.internal sackd[9985]: sackd: running

Confirm connections to the cluster are working using Slurm client commands such as sinfo and squeue. Here is example output from sinfo.

[root@ip-10-3-27-112 ~]# /opt/aws/pcs/scheduler/slurm-23.11/bin/sinfo
PARTITION AVAIL TIMELIMIT NODES STATE NODELIST
all up infinite 4 idle~ compute-[1-4]

You should also be able to submit jobs. For example, a command similar to this example would launch an interactive job on 1 node in the cluster.

/opt/aws/pcs/scheduler/slurm-23.11/bin/srun --nodes=1 -p all --pty bash -i