Auks (0.5.3) on RH7 issues with slurm passing the credential

[groucho64738]

Hi. I'm very very new at trying to get AUKS working and it's likely some configuration that I don't understand. This is using AUKS 0.5.3 on RH 7 and slurm 20.11.7. I've set up auksd on a management node and it appears to be functional. I set up our cluster head node and configured it to use the spank plugin for slurm and that appears to function as well.

• ssh into the cluster head node, klist shows my tickets, auks -p verifies, auks -a and -r work to load the keys into the cache. Watching the debug logs on the server show that my credential [email protected] accesses auks.
• submitting a job with slurm like: srun --auks=yes also shows accesses to auksd, so I feel that the plugin functions
• the compute node that the job runs on, however, only generates these messages: auks_krb5_stream: authentication failed : Software caused connection abort

I can sorta reproduce the issue on my head node if I do a kdestroy -A and try to ping auks, it fails, so it almost seems like my cluster node is not receiving the auks credential.

I'll attach my auks.conf and auks.acl files auksconf.txt auks.acl.txt

Thanks for any insight you have on this. Let me know if there's any other info I can provide that would help.

[fihuer]

Hi, Could you give us the Slurm plugin for Auks configuration file (probably /etc/slurm/plugstack.conf.d/auks.conf) and some log files (slurm and auksd would help) ?

btw, we are currently transferring the ownership of auks to another repo (https://github.com/cea-hpc/auks), could you re-open a Discussion/Issue over there ?