Dear Slurm developers, is there a way to have an strigger script run on RESUME or DRAINING state changes of a compute node? Best regards, Ulf Markwardt <ulf.markwardt@tu-dresden.de>
These event triggers are not available today, but should be easy to add.
Are there plans to add these even triggers? They would be very useful.
(In reply to Steve Ford from comment #2) > Are there plans to add these event triggers? They would be very useful. Not currently. The "5 - Enhancement" status in our bugzilla is used to capture outstanding enhancement requests, but unless we've specifically identified a target release there is no commitment from SchedMD to develop these. Absent development sponsorship, or additional customer demand, a lot of these unfortunately will go uncompleted. If you, or anyone else, is interested in what that process entails that please let me know. - Tim
Created attachment 13070 [details] Patch for bug 2991 For more information refer to the commit message in the patch file.
(In reply to Tim Wickberg from comment #3) > (In reply to Steve Ford from comment #2) > > Are there plans to add these event triggers? They would be very useful. > > Not currently. The "5 - Enhancement" status in our bugzilla is used to > capture outstanding enhancement requests, but unless we've specifically > identified a target release there is no commitment from SchedMD to develop > these. > > Absent development sponsorship, or additional customer demand, a lot of > these unfortunately will go uncompleted. If you, or anyone else, is > interested in what that process entails that please let me know. > > - Tim Dear Tim, Moe, and Steve, Here at BSC wanted to perform some actions whenever a node is out of and back to production. But we wanted to do so just at the moment of requesting so. Draining is desirable in front of drained, because if a node is not drained instantaneously then the trigger is raised far away from the draining start time. Resume is needed because a node could be on idle or allocated state without being back to prod, and we wanted to trigger only one time when a node goes back. Tell me what you think of this patch proposal. I hope to have this 2 flags available in next releases. Kind regards, - Carlos.
We would also really appreciate this feature. Any chance this can get merged soon?
Sorry for not responding sooner, we've been busy closing out the 20.02 release. I'll take a look at the proposed patch in a month or so for consideration for 20.11. Our freeze for major changes in 20.02 went in place in January, so I cannot consider it for 20.02. - Tim
(In reply to Tim Wickberg from comment #7) > Sorry for not responding sooner, we've been busy closing out the 20.02 > release. > > I'll take a look at the proposed patch in a month or so for consideration > for 20.11. Our freeze for major changes in 20.02 went in place in January, > so I cannot consider it for 20.02. > > - Tim Dear Tim, thanks for the information, and no worries about the 20.02 freeze. It's completely OK as long as the patch can be taken into account soon or later. Thank you very much, again, for your time. Kind regards, - Carlos.
Dear Tim, Do you have any advance on this? Thank you!
Hi there, I was about to submit a separate request for this, but since it's already been requested here, I'd like to voice our interest for this feature as well. Thanks! -- Kilian
Hi SchedMD, Would it be possible to get Carlos' patch merged in 22.05? Thanks! -- Kilian
Created attachment 26252 [details] Patch for draining/resume triggers, ported to 21.08.8 Here's a version of the patch ported to 21.08.8, that adds support for node draining and resume triggers. We have it deployed on our main system, and it seems to work as expected. Cheers, -- Kilian
Hope this can be merged soon.
Hi, This is implemented in 1b795f0325...83138422e9 series of commits in master branch. I'm going to mark this as resolved/fixed. Cheers, Carlos.
Hi Carlos, (In reply to Carlos Tripiana Montes from comment #24) > This is implemented in 1b795f0325...83138422e9 series of commits in master > branch. > > I'm going to mark this as resolved/fixed. This is great, thank you! Is there a chance this patchset could be back-ported to 22.05 as well? Maybe not merged in the tree (I understand it would be a new feature in an existing branch and may be an issue) but at least posted as a patch here? Thanks! Cheers, -- Kilian
I'm afraid we don't officially deliver backported patches when they are targeted to modern versions, but I'm fairly sure this would be pretty easy to do on your own. But, as always, we can't officially provide support for any derived consequences of such changes. Hope you'll understand. Cheers, Carlos.
