| jgilaber | #startmeeting watcher | 13:00 |
|---|---|---|
| opendevmeet | Meeting started Thu Oct 29 13:00:43 2026 UTC and is due to finish in 60 minutes. The chair is jgilaber. Information about MeetBot at http://wiki.debian.org/MeetBot. | 13:00 |
| opendevmeet | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 13:00 |
| opendevmeet | The meeting name has been set to 'watcher' | 13:00 |
| jgilaber | Hi! Who is around today? | 13:00 |
| rlandy | o/ | 13:00 |
| amoralej | o/ | 13:00 |
| winiciusallan[m] | o/ | 13:01 |
| jgilaber | While we gather, feel free to add topics to the agenda https://etherpad.opendev.org/p/openstack-watcher-irc-meeting#L29 | 13:01 |
| morenod | o/ | 13:01 |
| jgilaber | courtesy ping: dviroel sean-k-mooney chandankumar | 13:01 |
| dviroel | o/ | 13:01 |
| jgilaber | let's wait for another minute and we'll start | 13:02 |
| jgilaber | #topic Preemptible instances | 13:03 |
| jgilaber | winiciusallan[m], go ahead please | 13:03 |
| winiciusallan[m] | thanks jgilaber | 13:03 |
| winiciusallan[m] | so, last cycle, I proposed a new feature to Watcher to have preemptible instances | 13:04 |
| chandankumar | o/ | 13:04 |
| winiciusallan[m] | the feature wasn't merged -- only two new actions that was part of the whole effort | 13:04 |
| winiciusallan[m] | I'd like to resurrect that feature on this cycle, but I think this will require a few changes on the spec | 13:05 |
| winiciusallan[m] | I don't have enough free time to implement as-is | 13:05 |
| dviroel | the new actions: delete and shelve | 13:05 |
| winiciusallan[m] | Currently, we're proposing to use placement in the strategy, but this would require a lot of other changes (CDM, api etc) | 13:06 |
| winiciusallan[m] | by reducing the scope, I'd like to remove the placement part on this first iteration, then we could add it later if time permits | 13:06 |
| sean-k-mooney | o/ | 13:07 |
| winiciusallan[m] | I'm linking the spec in the agenda | 13:07 |
| dviroel | the placement part that you mention is about the priority sorting piece, right? | 13:07 |
| sean-k-mooney | removing the placmenet part woudl be ok but that also means removeing resouces: | 13:07 |
| dviroel | **resource:<name>** | 13:07 |
| sean-k-mooney | if you want to do flavor based sorting you need flavor:<name> or similar instead | 13:08 |
| dviroel | i think that doesn't block for adding it later if wanted to | 13:08 |
| winiciusallan[m] | yes. we use "resource:<name>" to sort by a given resource. we'll remove that and use only vcpu, ram, and disk | 13:08 |
| jgilaber | so instances would only be preempted by age, correct? | 13:08 |
| sean-k-mooney | ya that also works | 13:08 |
| sean-k-mooney | vcpu,ram,disk | 13:08 |
| sean-k-mooney | that avoid the depency on the placement enhancments | 13:09 |
| winiciusallan[m] | jgilaber: iirc we can use flavor information, not only age | 13:09 |
| jgilaber | ack, that sounds reasonable then | 13:09 |
| dviroel | the previous spec only mentions "age" and "resource:<name>" | 13:10 |
| dviroel | so the placement part would be replaced by flavor ones | 13:10 |
| dviroel | which is available in the model | 13:10 |
| dviroel | iirc | 13:10 |
| winiciusallan[m] | yes, exactly. there is an opened patch but I don't remember the last state https://review.opendev.org/c/openstack/watcher/+/1000170 | 13:11 |
| dviroel | I think that is fine | 13:11 |
| winiciusallan[m] | s/opened/open | 13:11 |
| amoralej | ordering by flavor ones, means vcpu, ram and disk, right? not by image name | 13:11 |
| dviroel | yeah | 13:11 |
| winiciusallan[m] | exactly amoralej | 13:11 |
| amoralej | sorry ^i mean not by flavor name | 13:11 |
| amoralej | i think that's a good starting point | 13:11 |
| winiciusallan[m] | what would be required from me in terms of moving it to this new cycle? | 13:12 |
| winiciusallan[m] | I need to propose a new patch to watcher-specs with the modifications? | 13:12 |
| jgilaber | winiciusallan[m], yes | 13:12 |
| dviroel | winiciusallan[m]: you need to propose it to 2027.1 approved path, yeah | 13:12 |
| jgilaber | and add a new row in the history table | 13:13 |
| winiciusallan[m] | alright. thanks! | 13:13 |
| winiciusallan[m] | is there anything to be changed in the blueprint? | 13:13 |
| jgilaber | thanks winiciusallan[m]. Any other comment/question on this topic? | 13:13 |
| dviroel | the blueprint is still open | 13:14 |
| dviroel | since we didn't complete it in previous cycle | 13:14 |
| dviroel | #link https://blueprints.launchpad.net/watcher | 13:14 |
| winiciusallan[m] | jgilaber: I'm good. thanks! | 13:15 |
| dviroel | thanks winiciusallan[m]++ | 13:15 |
| winiciusallan[m] | dviroel: gotcha. I can take a look at this later | 13:15 |
| jgilaber | ok, next topic | 13:15 |
| jgilaber | #topic stable/2025.1 backports | 13:15 |
| jgilaber | We merged a few patches there, but we still have plenty open | 13:16 |
| jgilaber | #link https://review.opendev.org/q/project:openstack/watcher+branch:stable/2025.1+status:open | 13:16 |
| jgilaber | do we want to merge anything else? | 13:16 |
| jgilaber | we should make the new release tomorrow ideally | 13:16 |
| jgilaber | to move to unmaintained | 13:16 |
| jgilaber | any thoughts? | 13:17 |
| jgilaber | if not, I can propose a new release today | 13:18 |
| dviroel | there is alot, hard to decide which ones are really important | 13:18 |
| dviroel | not sure if addin one or two more changes too much | 13:19 |
| jgilaber | I agree, so I think it's ok to leave it as is, but wanted to double check | 13:19 |
| amoralej | i agree | 13:20 |
| jgilaber | ack | 13:20 |
| jgilaber | I'll check after the meeting and make a release for all stable branches that have had changes | 13:20 |
| jgilaber | I don't know now if all had | 13:20 |
| dviroel | 2026.2 should have a few already | 13:21 |
| jgilaber | yes, 3 or 4 bugs fixes from what I can see | 13:22 |
| jgilaber | ok, anything else on this topic? | 13:23 |
| jgilaber | moving on | 13:23 |
| jgilaber | #topic Proposal to add ceph to the aetos job | 13:23 |
| jgilaber | We've had some bugs related to differences in reported info by different cinder drivers | 13:24 |
| jgilaber | so I've opened a patch to setup ceph in one of the jobs | 13:24 |
| jgilaber | I arbitrarely used the aetos one | 13:24 |
| jgilaber | #link https://review.opendev.org/c/openstack/watcher/+/1011633 | 13:24 |
| jgilaber | any objection to that? | 13:24 |
| jgilaber | the job seems to work well | 13:25 |
| dviroel | the prometheus datasource is deprecated, so it is aethos or gnocci | 13:25 |
| dviroel | s/aethos/aetos | 13:26 |
| dviroel | ok for me, but I didn't check the job and patch yet | 13:26 |
| amoralej | sounds good, i didn't check the patch, but i think that'd be good | 13:26 |
| jgilaber | ack sure, it's not an urgent review, just a nice to have | 13:26 |
| jgilaber | I think we can move on and discuss in the patch | 13:27 |
| jgilaber | #topic Proposed spec to change strategies updates | 13:27 |
| jgilaber | this one if from amoralej | 13:27 |
| dviroel | jgilaber: ack, thanks | 13:27 |
| jgilaber | #link https://review.opendev.org/c/openstack/watcher-specs/+/1012696 | 13:27 |
| amoralej | I sent that spec to propose some modifications related to current strategies update process | 13:27 |
| amoralej | we discussed it in the PTG | 13:28 |
| dviroel | i think that goes in the direction that we agreed in the ptg | 13:28 |
| amoralej | I already got review from dviroel but i'd like to get more reviews so that i propose an update | 13:28 |
| amoralej | I'll detail more an implementation for data migration for the parameters spec | 13:29 |
| dviroel | ++ | 13:29 |
| jgilaber | I'll add it to my review list, but I doubt I'll get to that this week | 13:29 |
| amoralej | ok, thanks jgilaber | 13:29 |
| jgilaber | ok, let's move to reviews | 13:30 |
| jgilaber | #Reviews | 13:30 |
| jgilaber | #link https://review.opendev.org/q/project:openstack/watcher-specs+status:open | 13:30 |
| jgilaber | we have 6 specs open, with one more on the way | 13:30 |
| dviroel | I added as a reminder | 13:30 |
| jgilaber | thanks dviroel | 13:30 |
| dviroel | to each one review each others specs | 13:31 |
| dviroel | :) | 13:31 |
| dviroel | as usual | 13:31 |
| dviroel | nothing specific | 13:31 |
| jgilaber | we should start prioritizing them to avoid having to scramble close to M! | 13:31 |
| jgilaber | M1 | 13:31 |
| dviroel | yep | 13:31 |
| jgilaber | ok, moving to next review | 13:32 |
| jgilaber | #link https://review.opendev.org/c/openstack/watcher-tempest-plugin/+/1012005 | 13:32 |
| amoralej | M1 is final date for specs? | 13:32 |
| jgilaber | I'm looking at the docs, I might have mixed up with M" | 13:33 |
| jgilaber | M2 | 13:33 |
| dviroel | amoralej: usually m2 | 13:33 |
| dviroel | but we need to consider that we also have holidays in the middle,so we should focus on getting them merged asap yeah | 13:33 |
| amoralej | ok, good, i was thinking in the transformers one :) | 13:33 |
| dviroel | :) | 13:33 |
| amoralej | yep, the sooner the better | 13:33 |
| jgilaber | indeed, I mixed them up, sorry | 13:34 |
| dviroel | "Unskip test_list_with_limit now that bug 2134046 is fixed" | 13:34 |
| jgilaber | spec feature freeze is the week Jan 11 - Jan 15 | 13:34 |
| jgilaber | back to morenod's patch | 13:34 |
| jgilaber | is that bug fix backportable? | 13:34 |
| jgilaber | if not we should guard the test execution for stable branches | 13:34 |
| morenod | tests have passed on all the versions | 13:35 |
| morenod | https://zuul.opendev.org/t/openstack/buildset/bb272d1d912e405b97c6494e87a47136 | 13:35 |
| jgilaber | but that problem was not consistent right? | 13:35 |
| dviroel | it merged in feb 2026 | 13:35 |
| dviroel | so it would be good to backport at least for 2025.2 | 13:36 |
| jgilaber | so we could see ocasional failures in jobs for 2025.2 | 13:36 |
| dviroel | since 2025.1 is going unmaintained, and we could remove it ci job from the plugin | 13:36 |
| jgilaber | dviroel, yes, we need to do that as well | 13:36 |
| dviroel | i would hold yes in favor of the backport | 13:37 |
| jgilaber | +1 | 13:37 |
| jgilaber | other thoughts/questions on this patch? | 13:37 |
| dviroel | amoralej: would you cherry pick this one? :) | 13:38 |
| amoralej | sorry, i was side tracked | 13:38 |
| amoralej | https://review.opendev.org/c/openstack/watcher/+/977786 ? | 13:39 |
| jgilaber | yes | 13:39 |
| dviroel | yeah | 13:39 |
| amoralej | yep | 13:39 |
| jgilaber | ok, looks like we're sorted here | 13:40 |
| amoralej | i need to check which branch already have it | 13:40 |
| amoralej | but i agree on removing the 2025.1 from the gate btw | 13:40 |
| jgilaber | it looks like it was merged in gazpacho, so we only need to backport it to 2025.2 | 13:40 |
| amoralej | yep | 13:41 |
| jgilaber | thanks amoralej | 13:41 |
| jgilaber | #topic Bugs | 13:41 |
| jgilaber | we have a few bugs today to triage | 13:41 |
| jgilaber | #link https://bugs.launchpad.net/watcher/+bug/2170351 | 13:42 |
| dviroel | amoralej: i think that you already hit this one ^? | 13:42 |
| dviroel | workload_stabilization with host_choice=fullsearch | 13:43 |
| amoralej | yes, i hit it when running functional tests | 13:43 |
| jgilaber | do we have another report for this then? | 13:43 |
| amoralej | it happens when there is more that one node proposed as candidate for migrations depending on the metrics | 13:43 |
| amoralej | no, i realized it was reported | 13:44 |
| amoralej | so i didn't open a new one | 13:44 |
| jgilaber | ok, so then we can set it to confirmed | 13:44 |
| dviroel | i say a comment in functional tests framework | 13:44 |
| dviroel | about that | 13:44 |
| jgilaber | and medium? since there seems to be a workaround | 13:45 |
| dviroel | there is fix attached, but not proposed? | 13:45 |
| amoralej | yes | 13:45 |
| amoralej | actually, i'm not sure that's the best fix | 13:45 |
| amoralej | i have it in my todo list, but didn't reach it | 13:45 |
| dviroel | the workaround is to use a different destination selection option | 13:45 |
| dviroel | not really the same thing | 13:46 |
| dviroel | but yeah, we can set as medium, it raises an exception and audit fails right | 13:46 |
| amoralej | medium seems fine | 13:46 |
| jgilaber | ok, thanks | 13:46 |
| jgilaber | #link https://bugs.launchpad.net/watcher/+bug/2172449 | 13:46 |
| dviroel | jgilaber: i guess that you are updating them already, right? | 13:46 |
| jgilaber | next one reported by morenod | 13:47 |
| jgilaber | yes, I've already did | 13:47 |
| dviroel | tks | 13:47 |
| morenod | I keep testing audit scope based on the comments from amoralej, and I am getting new errors | 13:47 |
| dviroel | "instance_metadata" doesn't work, is that it? | 13:48 |
| dviroel | in scope | 13:48 |
| morenod | yes | 13:48 |
| dviroel | damn | 13:48 |
| amoralej | actually, i didn't know that option even existed tbh ... | 13:48 |
| dviroel | yeah, audit scope is poorly documented and tested | 13:49 |
| amoralej | we have received question about "marking instances to not be migrated by watcher" ^ may be an option if it works | 13:49 |
| dviroel | GET /v1/data_model?data_model_type=compute&audit_uuid=<uuid> returns a 500? | 13:49 |
| dviroel | only for this specific audit? or any failed? | 13:50 |
| amoralej | as per the bug report, i guess any | 13:50 |
| amoralej | as it seems to be happening when creating the scoped model | 13:50 |
| dviroel | ack, we may need a new bug report, but we can check that in details in the code | 13:51 |
| morenod | I have tested with dummy, but yes, it will be probably fails with any strategy | 13:51 |
| dviroel | Importance? | 13:52 |
| dviroel | i think that we should fix this asap | 13:52 |
| amoralej | mmm actually the check_optimize_metadata is strategy option, so it may be only for this, dunno | 13:52 |
| dviroel | from basic_consolidation yeah | 13:53 |
| dviroel | this is probably a bad design | 13:53 |
| jgilaber | I would need to dig deeper, but this looks like high to me | 13:54 |
| amoralej | so, scoping by metadata is only intended to work with that? that's uggly | 13:54 |
| amoralej | i mean only with that strategy | 13:54 |
| dviroel | yeah, that may require major discussion on how to proper fix | 13:54 |
| dviroel | i need to look in details in the code | 13:54 |
| dviroel | and we can agree with a fix | 13:54 |
| dviroel | high? | 13:55 |
| dviroel | someone wants to be assignee of this one? | 13:55 |
| amoralej | we can implement a fix to avoid that particular issue but it probably deserves discussion about the feature itself | 13:56 |
| dviroel | yeah | 13:56 |
| dviroel | but we need to discuss the entire issue before merging the fix i guess | 13:56 |
| jgilaber | sounds good, triaged and high? | 13:56 |
| dviroel | I can assigne to myself, but don't plan to work right now on it. | 13:56 |
| amoralej | +1 | 13:56 |
| dviroel | jgilaber: +1 | 13:57 |
| jgilaber | I think it's better to leave it unassigned if we can't work on it | 13:57 |
| amoralej | dviroel, or keep it unasign so that whoever get some cycles take it | 13:57 |
| dviroel | ack | 13:57 |
| dviroel | sound good too | 13:57 |
| jgilaber | ok, done let get through one more quickly | 13:57 |
| jgilaber | #link https://bugs.launchpad.net/watcher/+bug/2172453 | 13:58 |
| dviroel | we can revisit the High ones time from time | 13:58 |
| amoralej | good point | 13:58 |
| amoralej | we could do that a specific section in mtgs | 13:58 |
| dviroel | about the #2172453 | 13:59 |
| dviroel | if the maintenance_node doesn't exist in the model | 13:59 |
| dviroel | which can be a removed node, or exclude via scope | 14:00 |
| dviroel | should the strategy fails> | 14:00 |
| dviroel | ? | 14:00 |
| amoralej | mmm for me that's expected behavior | 14:00 |
| morenod | I would say an empty action plan | 14:01 |
| jgilaber | I'm undecided on whether it should produce and action plan or an error | 14:01 |
| amoralej | but succeeded would imply the vms have been moved and the hw emptied | 14:01 |
| amoralej | which is not | 14:01 |
| jgilaber | Ideally we would reject that audit with a 400, but that would be quite hard I think | 14:01 |
| amoralej | yes, ideally it'd be rejected at creation | 14:01 |
| dviroel | that's a fair discussion, mainly because strategies implement behavior differently | 14:01 |
| amoralej | but we don't check values of parameters | 14:01 |
| amoralej | let's think a user creates a host_matintenance: from src_node to dest_node. There are no VMS. It goes to SUCCEEDED with empty AP | 14:02 |
| amoralej | that's fine | 14:02 |
| dviroel | we could agree on how they should behave for specific scenarios like that. and document it | 14:02 |
| amoralej | but we can not have the same behavior with a case that is not emptying the host imo | 14:03 |
| dviroel | SUCCEEDED -> error_messge: maintenance_node could not be found | 14:03 |
| dviroel | :) | 14:03 |
| amoralej | unless we indicate in the efficacy indicators that goal is achieved at 0% | 14:03 |
| amoralej | and even with that, i'd be careful, tbh | 14:04 |
| jgilaber | should we improve the error reporting in this case? | 14:04 |
| jgilaber | we're over time, do we want to leave it as is and revisit it next week? | 14:05 |
| amoralej | using status_message may be an option, right | 14:05 |
| dviroel | so in the end, the algorithym for the optimization didn't run due to a missing node, so I would say that it should be a FAILED run | 14:05 |
| dviroel | yeah, lets revisit | 14:06 |
| jgilaber | +1 to that, but let's check again next week | 14:06 |
| dviroel | thanks for reporting those morenod++ | 14:06 |
| jgilaber | #topic Volunteer to chair next meeting | 14:06 |
| jgilaber | any volunteer? | 14:06 |
| amoralej | i can chair | 14:06 |
| dviroel | amoralej++ | 14:06 |
| jgilaber | thanks amoralej | 14:06 |
| jgilaber | #topic Open discussion | 14:06 |
| jgilaber | any last minute topic? | 14:06 |
| jgilaber | ok, closing then | 14:07 |
| jgilaber | thanks everyone! | 14:07 |
| jgilaber | #endmeeting | 14:07 |
| opendevmeet | Meeting ended Thu Oct 29 14:07:38 2026 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 14:07 |
| opendevmeet | Minutes: https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-10-29-13.00.html | 14:07 |
| opendevmeet | Minutes (text): https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-10-29-13.00.txt | 14:07 |
| opendevmeet | Log: https://meetings.opendev.org/meetings/watcher/2026/watcher.2026-10-29-13.00.log.html | 14:07 |
| dviroel | thanks for chairing jgilaber | 14:07 |
| opendevreview | chandan kumar proposed openstack/watcher-dashboard master: [DNM] reproduce https://bugs.launchpad.net/openstack-cyborg/+bug/2172447 https://review.opendev.org/c/openstack/watcher-dashboard/+/1012875 | 14:08 |
| opendevreview | Alfredo Moralejo proposed openstack/watcher stable/2025.2: Check Audit state in the decision-engine before starting the execution https://review.opendev.org/c/openstack/watcher/+/1012876 | 14:09 |
| dviroel | chandankumar: i think that you linked the wrong bug LOL | 14:19 |
| chandankumar | yes, I did | 14:20 |
| chandankumar | let's wait for the CI run to finish then I can fix it | 14:20 |
| dviroel | sure | 14:22 |
| opendevreview | Merged openstack/watcher stable/2026.1: Skip unscheduled volumes in Cinder collector https://review.opendev.org/c/openstack/watcher/+/1007828 | 15:06 |
| jgilaber | dviroel, can you take a look at https://review.opendev.org/c/openstack/watcher/+/1007830? It's the last backport for this bug | 15:09 |
| jgilaber | and then I can update the release hashes | 15:09 |
| dviroel | ack, in a meeting now, but will take a look | 15:09 |
| dviroel | should be wuick | 15:09 |
| jgilaber | sure, thanks! | 15:09 |
| opendevreview | Merged openstack/watcher stable/2025.2: Skip unscheduled volumes in Cinder collector https://review.opendev.org/c/openstack/watcher/+/1007830 | 17:01 |
| opendevreview | Joan Gilabert proposed openstack/watcher-specs master: Add spec for OpenStackSDK migration 2027.1 https://review.opendev.org/c/openstack/watcher-specs/+/1005750 | 17:10 |
| opendevreview | Merged openstack/watcher stable/2025.2: Check Audit state in the decision-engine before starting the execution https://review.opendev.org/c/openstack/watcher/+/1012876 | 17:59 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!