[slurm-users] logging question

Derek Yarnell derek at umiacs.umd.edu
Wed Jan 10 07:57:42 MST 2018


Hi,

So we have been trying to track down GRES/cgroup issues that we have
been having with multi-GPU space sharing.  We were running 17.02.2 and
upgraded to 17.02.9 and it looks like the some of the messages were
changed from debug2 to info level based on the diff of
src/plugins/task/cgroup/task_cgroup_devices.c.

@@ -431,7 +426,7 @@ extern int
task_cgroup_devices_create(stepd_step_rec_t *job)
                 * are supposed to be allowed by default.
                 */
                for (k = 0; k < allow_lines; k++) {
-                       debug2("Default access allowed to device %s for
step",
+                       info("Default access allowed to device %s",
                             allowed_dev_major[k]);
                        xcgroup_set_param(&step_devices_cg, "devices.allow",
                                          allowed_dev_major[k]);


What I would expect is that these to still be output when we have higher
level specified.  From the man page for debug5 it says, Log errors and
verbose informational messages and even more debugging messages.

SlurmdDebug = 5
SlurmdLogFile = /var/log/slurm/slurmd.log

We have the above config set.  But we no longer get the messages about
the Default access allowed to devices.  Is this expected behavior for it
not to capture info messages at debug5?

Thanks,
derek

-- 
Derek T. Yarnell
Director of Computing Facilities
University of Maryland
Institute for Advanced Computer Studies



More information about the slurm-users mailing list