15:00:18 #startmeeting ironic 15:00:19 Meeting started Mon Aug 17 15:00:18 2020 UTC and is due to finish in 60 minutes. The chair is TheJulia. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:19 o/ 15:00:20 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 15:00:22 The meeting name has been set to 'ironic' 15:00:25 o/ 15:00:28 o/ 15:00:32 Good morning Ironic! 15:00:51 Our meeting agenda this week can be found on the wiki. 15:00:52 #link https://wiki.openstack.org/wiki/Meetings/Ironic#Agenda_for_next_meeting 15:00:57 o/ 15:01:02 o/ 15:01:14 \o 15:01:15 o/ 15:01:16 ERROR: Insufficent [\o|o/] : More coffee required 15:01:19 :) 15:01:29 #topic Announcements / Reminder 15:01:38 o/ 15:01:40 o/ 15:02:07 We had several of our components have their second release of the cycle last week. We have a couple others still in the final pipeline as they were not approved last week. 15:02:08 o/ 15:02:18 Aside from that, all quiet. Does anyone have anything to announce or remind us of? 15:03:32 * TheJulia hears https://www.youtube.com/watch?v=K8E_zMLCRNg :) 15:03:39 (crickets) 15:03:49 a reminder: 15:04:05 after creating a bugfix branch, do NOT approve the constraints changes 15:04:18 we need to fix the release tooling not to propose them at all, they're broken 15:04:18 did we forget to release something ? 15:04:31 dtantsur: yeah 15:04:45 dtantsur: definitely need to fix the tooling then, they are too easy to auto-pilot approve on 15:04:47 for now manually update the patches to replace X.Y with victoria 15:04:49 yep 15:04:57 I had to hold my hand several times already 15:04:59 Do we have any to revert? 15:05:19 wonder if there is anywhere else we can record that info, cuz yeah, too easy to approve :-( 15:05:37 TheJulia: we need to fix one, inspector or bifrost, I don't remember 15:05:56 Well, we can't even update them manually now because the victoria branch does not exist yet 15:06:09 joy! 15:06:12 btw the release tool now has a new option for intermediate-branch https://review.opendev.org/745918 15:06:13 patch 745918 - releases - Add intermediate-branch arg to new-release script (MERGED) - 1 patch set 15:06:22 Okay, well, we can carry on! 15:06:45 We had no action items from our last meeting, so if there is no objection we can proceed to the whiteboard 15:07:49 * TheJulia hears no objections 15:07:53 #topic Review subteam status reports 15:08:00 #link https://etherpad.openstack.org/p/IronicWhiteBoard 15:08:02 TheJulia: the victoria branch does not exist, but the constraints link does 15:08:11 dtantsur: ack 15:08:18 i.e. https://releases.openstack.org/constraints/upper/victoria is already valid and should be used 15:08:25 Starting around line 310 on the whiteboard 15:08:28 (it redirects to master) 15:11:40 Looks like the Standalone auth item is done? 15:11:51 I hope so 15:13:23 I see the WSME related changes are continuing to move forward although they seem to be fighting CI headaches 15:14:52 Looks like duplicate job changes have kind of stalled? 15:16:23 * TheJulia sees dtantsur add an update for the section she was about to ask about 15:16:28 heh 15:16:49 we should maybe decide if we want to keep the name of the image used in the jobs, tinyipa or dib 15:17:32 rpioso: arne_wiebalck: rpittau: Any new updates on the redfish interop profiles. (This is where dating notes helps a lot) 15:17:55 rpittau: that is a good question 15:18:08 TheJulia: no news, I think last ujpdate is from arne_wiebalck on July 28 15:18:17 rpittau: ack 15:18:41 I guess once arne_wiebalck is mentally back we'll need to discuss next steps 15:19:18 I think rpioso is giving an update :) 15:19:46 wesome 15:19:48 awesome 15:21:11 awesome 15:21:21 rpittau, TheJulia: Done 15:21:22 Well, is everyone good to move onward? 15:21:30 rpioso: thanks! 15:21:37 TheJulia: yw 15:22:54 Everyone good to proceed to priorites for the week? 15:23:28 let's 15:23:41 #topic Deciding on priorities for the coming week 15:23:52 #link https://etherpad.opendev.org/p/IronicWhiteBoard 15:24:01 Starting at line 159 15:24:07 First up, clean up the merged items! 15:25:48 Okay, I've staged some items at line 238 through 262. Any objections to adding them? 15:25:56 Any additional items people would like to see added? 15:26:21 * dtantsur has an empty list this time 15:26:57 should be ok 15:28:22 Does the list look good at this time? 15:29:14 yep 15:30:06 So onward to discussion! Although one item was raised for discussion in the list of items to review. :) 15:30:34 We sort of briefly visited it last week, so we can get to that again during Open Discussion 15:30:40 #topic Discussion 15:30:56 One item this week, and largely raising, yet another question, when shall we have another midcycle? 15:32:27 * TheJulia wonders if youtube would grant her another random crickets video 15:32:28 victoria will be released on October 14th, so maybe in 2 weeks ? 15:32:38 too short notice? 15:32:44 I think that is about perfect 15:33:16 Any volunteers to create a poll? 15:33:22 that should be the 1st week of September 15:33:27 I can do it 15:33:31 We can start an etherpad this week as well for ideas 15:34:21 Awesome! 15:34:25 sounds good 15:34:35 Anyone disagree? 15:34:55 or have any other thoughts on the topic before we jump to Open Discussion? 15:35:14 arne_wiebalck: I'm guessing Baremetal SIG has nothing this week to discuss at the moment, please let me know if this is not the case 15:37:28 * TheJulia hears crickets 15:37:32 #topic Open Discussion 15:38:15 So one item to be revisited I guess for open discussion is https://etherpad.opendev.org/p/IronicWhiteBoard line # 169. 15:38:50 Which kind of brings back the question of "How would an operator interact?", I'm wondering if anyone has more thoughts on this? 15:40:24 * TheJulia senses she needs to brew coffee for the IRC channel this morning 15:40:50 arne_wiebalck seems ok with populating properties :) 15:41:19 but the overhead for large-scale cloud, I don't have the precise stats of the overhead 15:42:39 A lot of CPU time and api calls gets consumed keeping nova in sync with ironic. So adding more work to that series of events requires careful consideration 15:42:57 I guess I was also wondering if it would be data operators would ever have to change 15:43:02 or would it be entirely inspector managed 15:44:12 How are these acclerator devices in node supposed to be used? To create resource classes subsequently... 15:44:50 Great question 15:44:58 they are expected to be used for scheduling 15:45:23 * TheJulia wonders if so, do we publish it directly... 15:45:44 kaifeng: through capabilities? 15:45:48 That would allieviate most of my scale concern 15:45:48 JayF: Thank you for the great information the other day regarding Hardware Managers. They seem to be more simple than I had expected, which is good. One thing that I don't fully understand, though, is how they are loaded. Is it supposed to be that the code is just placed into the hardware_managers directory? I tried that, and tried clearing the cache directory, but the new ones still don't seem to be loaded. 15:46:11 for precise scheduling, user can create accel profile in the flavor, it can't be satisfied by trait, it's too simple. 15:46:55 dking: it has to be installed so the package setup metadata can point to it as a distinct endpoint... if memory serves 15:47:16 No, not going through capabilities, actually it's barely used in the resource class based scheduling 15:47:49 TheJulia: Okay. I didn't find any information on actually installing the new hardware managers, only info on how to create them and how they work. 15:48:28 dking: are you following the examples? https://opendev.org/openstack/ironic-python-agent/src/branch/master/examples 15:48:38 kaifeng: is it something like in cyborg? https://docs.openstack.org/api-guide/compute/accelerator-support.html 15:48:59 dking: this package: https://opendev.org/x/proliantutils/src/branch/master/setup.cfg#L31 has an example of the entrypoint creation. When installed as a python package, the data is updated and ironic_python_agent finds it as a result 15:49:20 dtantsur: Yes, those were the ones I was trying to mimic.... Oh, I see now that readme mentions setup.cfg. Let me check that. 15:49:22 dking: funny thing is, that package also has other code totally unrelated to the hardware manager embedded in it (in other words, its a good example) 15:49:36 stendulker: it is, we are pairing with virtual machine in accelerator support 15:49:56 our examples also have setup.cfg https://opendev.org/openstack/ironic-python-agent/src/branch/master/examples/business-logic/setup.cfg 15:49:57 kaifeng: ok 15:50:12 dtantsur: \o/ 15:50:21 Regarding the spec at https://review.opendev.org/#/c/740721/12 I left a comment at the very end about ways to improve security for configuration file storage. Can anyone take a look and see if that's OK direction to go or suggest alternatives? 15:50:21 patch 740721 - ironic-specs - Operationally efficient & fast-track provisioning - 12 patch sets 15:51:00 ajya: I should be able to look later today or tomorrow 15:51:07 thanks, TheJulia 15:51:55 I think the scaling issue can't be addressed by introducing a new field, as a scheduling metric, it has to be collected by the data consumer AIUI, 15:52:40 Merged openstack/ironic-inspector master: Fix URL parsing to enable redfish_address matching https://review.opendev.org/745972 15:52:47 kaifeng: There are some accel discovery done by IRMC driver https://github.com/openstack/ironic/blob/master/ironic/conf/irmc.py#L83-L100 15:52:48 kaifeng: as long as we're getting the field in the same get request from ironic, that is about all that really can be done... I think. 15:53:26 kaifeng: the additional dictionary handling in nova-compute wouldn't be great but I guess it would be necessary 15:55:07 Merged openstack/networking-generic-switch master: Generate key on PEM format https://review.opendev.org/739186 15:55:10 stendulker: thanks for the link, will take a look, looks like something we have in the pci_devices plugin from inspector 15:55:26 stendulker: I think where what kaifeng is proposing is different is wanting to take it further to facilitate scheduling data population somehow 15:55:45 wow, two changes merged. I guess CI maybe might be a little happier today then it was at first look? 15:56:03 Well everyone, we're about at time. 15:56:09 If there is nothing else we can end the meeting 15:56:20 kaifeng, TheJulia : Yes, I understand. I was pointing that we would have to consolidate them in a fashion that would be usable. I think even ilo has some inspection data around gpus. 15:56:38 stendulker: ahh! 15:56:39 ilo driver 15:56:45 stendulker: I see, makes sense 15:57:47 i think its just dumped in node properties as these were done when capabilities was getting used for scheduling 15:58:22 :\ 15:58:35 This seems like a larger conundrum that needs to be visited 15:58:49 Anyway, Thanks everyone! 15:59:21 thanks1 15:59:32 #endmeeting