Hi Forum, First, cluster info: Debian Buster (armbian) on arm64 architecture -- really nothing too fancy going on here.
I've built slurm from source with PMIX with the following steps: - Configure openmpi with slurm and internal pmix -- works as intended with mpirun/mpiexec - install debian package "libpmix-dev" - Configure slurm pointing at pmix dev library, and munge. I copied the ".service" files from the build directory into my /etc/systemd/system/. I modified these files to point to the right configure file and such. The only thing I'm not sure about is where to point the EnvironmentFile in that unit file. Now, here's the question: I can run # systemctl enable slurmctld No errors are printed and systemctl status says it's just inactive I then start the service with # systemctl start slurmctld No errors are printed and the controller can successfully communicate with the compute nodes Now on restart, the service does not automatically start because of the following error: slurmctld.service - Slurm controller daemon Loaded: loaded (/etc/systemd/system/slurmctld.service; enabled; vendor preset: enabled) Active: failed (Result: exit-code) since Fri 2020-12-11 21:32:47 GMT; 8min ago Process: 455 ExecStart=/usr/local/sbin/slurmctld -D $SLURMCTLD_OPTIONS (code=exited, status=1/FAILURE) Main PID: 455 (code=exited, status=1/FAILURE) Dec 11 21:32:47 ApacheHead systemd[1]: Started Slurm controller daemon. Dec 11 21:32:47 ApacheHead systemd[1]: slurmctld.service: Main process exited, code=exited, status=1/FAILURE Dec 11 21:32:47 ApacheHead systemd[1]: slurmctld.service: Failed with result 'exit-code'. Despite these errors, I can still start the service with the systemctl start command. Also, running the actual slurmctld command found in sbin runs correctly with no critical errors. I've tried to look into this, but can't seem to find too much on this problem for slurm or for system processes in general. Any ideas? Thanks, ~Avery Grieve They/Them/Theirs please! University of Michigan
