Custom Query (101 matches)

related to recent changes in the pyslurm library that jobmond.py uses to interface with Slurm. It was recently re-worked to bring it up to date with the most recent Slurm API and along the way the seem to have changed some of their data structures. Below is a patch I used to get it to run at our site. I hope this helps.

--- jobmond/jobmond.py  2014-01-20 09:24:08.000000000 -0700
+++ /usr/local/sbin/jobmond.py  2014-12-16 17:22:00.501223234 -0700
@@ -1306,7 +1306,7 @@

        for node, attrs in slurm_nodes.items():

-            ( num_state, name_state ) = attrs['node_state']
+            name_state = attrs['node_state']

            if name_state == 'DOWN':

@@ -1371,7 +1371,7 @@
            else:
                ppn = min_cpus

-            ( something, status_long ) = self.getAttr( attrs, 'job_state' )
+            status_long = self.getAttr( attrs, 'job_state' )

            status = 'Q'

Thanks,
Mike Robbert
HPC Engineer
Colorado School of Mines

seems a old bug has resurfaced: down/offline nodes are no longer (correctly) reported

I'm using the current version 0.3 and if there are no jobs running on the cluster, the pie chart does not show up (broken link icon).

Thanks.