Segmentation fault on pyslurm.node() about pyslurm HOT 11 CLOSED

Yep, this is has been raised by Jonathon Anderson on the Slurm mailing list and I'm currently working on it. How many nodes in your cluster ? The reason I ask is so I can recreate the issue.

from pyslurm.

I'll have to check but we're above 40 nodes at that point.
Aurélien

On Dec 9, 2015, at 2:28 PM, Gingergeek [email protected] wrote:

Yep, this is has been raised by Jonathon Anderson on the Slurm mailing list and I'm currently working on it. How many nodes in your cluster ? The reason I ask is so I can recreate the issue.

—
Reply to this email directly or view it on GitHub #52 (comment).

from pyslurm.

Thanks, sometime ago, I changed the code so I could track and manage the change in data (various classes) from the Slurm API so that looped monitoring programs could be used. Clearly I have a memory issue here but this was not showing up on my small test VM.

from pyslurm.

Aurélien,
I have traced a possible cause to a defined node in the config which does not truly exist. The Slurm API returns a record for it but with no data content. A quick check for a node name of NULL in the record was all that was necessary, does this match your configuration ?

Mark

from pyslurm.

Mark,
I'll have to get back to you on this. Are you asking if one of the node has a name 'NULL'?

Aurélien

On Dec 9, 2015, at 4:17 PM, Gingergeek [email protected] wrote:

Aurélien,
I have traced a possible cause to a defined node in the config which does not truly exist. The Slurm API returns a record for it but with no data content. A quick check for a node name of NULL in the record was all that was necessary, does this match your configuration ?

Mark

—
Reply to this email directly or view it on GitHub #52 (comment).

from pyslurm.

Aurélien,
It was a statement that we have traced the problem to a bug in PySlurm and not handling a node record with an empty nodename entry. I think this is where nodes have been entered in the slurm config but do not actually exist in the hosts file or DNS. My question was if it is possible that you have nodes in your config that do not actually exist ?

Mark

from pyslurm.

Aurélien,
If you pull the latest 14.11.5 the patch as discussed previously has been committed. Please let me know if this now works and close the ticket if you are happy.

Mark

from pyslurm.

Maybe... I do not have a complete understanding of how the cluster is managed, unfortunately.
Aurélien

On Dec 10, 2015, at 1:17 AM, Gingergeek [email protected] wrote:

Aurélien,
It was a statement that we have traced the problem to a bug in PySlurm and not handling a node record with an empty nodename entry. I think this is where nodes have been entered in the slurm config but do not actually exist in the hosts file or DNS. My question was if it is possible that you have nodes in your config that do not actually exist ?

Mark

—
Reply to this email directly or view it on GitHub.

from pyslurm.

Aurélien,
This should now be resolved with the latest pyslurm-14.11.5 commits. So please test and close the ticket if it is fixed.

Mark

from pyslurm.

Aurélien,
Have you had chance to test and confirm it is now resolved ?

Mark

from pyslurm.

Closing this as resolved

from pyslurm.