Ticket 10195

Summary: Power saving and node weight
Product: Slurm Reporter: OCF Support <support>
Component: slurmrestdAssignee: Director of Support <support>
Status: RESOLVED DUPLICATE QA Contact:
Severity: 4 - Minor Issue    
Priority: --- CC: dbartkiewicz, nrini
Version: 20.02.5   
Hardware: Linux   
OS: Linux   
Site: OCF Slinky Site: ---
Alineos Sites: --- Atos/Eviden Sites: ---
Confidential Site: --- Coreweave sites: ---
Cray Sites: --- DS9 clusters: ---
Google sites: --- HPCnow Sites: ---
HPE Sites: --- IBM Sites: ---
NOAA SIte: --- NoveTech Sites: ---
Nvidia HWinf-CS Sites: --- OCF Sites: Schlumberger
Recursion Pharma Sites: --- SFW Sites: ---
SNIC sites: --- Tzag Elita Sites: ---
Linux Distro: --- Machine Name:
CLE Version: Version Fixed:
Target Release: --- DevPrio: ---
Emory-Cloud Sites: ---

Description OCF Support 2020-11-11 02:38:28 MST 
The slurm power saving feature has been turned on, so that idle nodes are powered off when not in use. It has been noticed that new jobs are scheduled to nodes that have recently been powered on, rather than nodes that should be prioritised due to the node "Weight". It looks like there are already bugs files in the slurm bugtracker (9548 & 9734) , currently marked as "under consideration".
Comment 1 OCF Support 2020-11-11 02:39:44 MST 
Before the customer started using the powersave feature, all other things being equal, jobs would be scheduled to nodes[001-051] in preference to gnodes[001-016], as the Weight was lower.

After turning on the powersave feature, the weight appears to be ignored. Everything else works as expected (eg specific feature or resource requests).
Comment 2 Nate Rini 2020-11-11 10:12:00 MST 
I'm going to mark this as a duplicate:

(In reply to Broderick Gardner from bug#9734 comment #1)
> Yes, as it is currently designed, nodes in a powered-down state are in a
> lower tier altogether, making weights useless for cloud. I'm looking into
> what the scope for an enhancement would be in this regard.

Please take a look at bug#9734 for continuing updates.

*** This ticket has been marked as a duplicate of ticket 9734 ***