Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-12-05
20:14:16 efried but I need to relook at how ugly it is.
20:14:45 mriedem i've abandoned
20:14:51 mriedem people can cherry pick it downstream if they really need it
20:14:58 efried okay
20:15:38 efried mriedem: I guess you're going to rebase and clean up the other one now, so I should wait to review?
20:16:01 efried yeah
20:16:14 mriedem i'm not going to get to that today but it won't be much different
20:16:25 mriedem probably crank it out in the wee morn
20:16:34 efried I didn't open any of the files, but the commit message at least needs an overhaul.
20:17:12 mriedem ubetcha
20:17:48 efried okay, let's see if I can think clearly enough to get back to rg/rp mappings...
20:18:46 efried mriedem: if you're feeling magnanimous, you could hit the bottom three patches here, fairly simple: https://review.opendev.org/#/c/696418/
20:19:34 mriedem oh god request groups
20:19:45 mriedem i think melwitt said she wanted to learn more about this part of the code
20:19:49 efried hahaha
20:20:17 efried In my IRC client, melwitt shows up a nice grasshopper green. Coincidence?
20:21:23 mriedem as in the kung fu grasshopper pupil green?
20:21:45 mriedem or the "oh god request groups, i think i'm going to be sick" green
20:32:21 mriedem gibi: a few nits in https://review.opendev.org/#/c/696245/ if you want to start a FUP patch for that series
20:34:17 openstackgerrit Eric Fried proposed openstack/nova master: WIP: Use provider mappings from Placement (mostly) https://review.opendev.org/696992
21:00:27 efried meeting now
21:14:10 efried mriedem: The vote in the meeting was unanimous: do a spec for the ironic owner filter
21:14:17 efried (unanimous: both of us)
21:14:54 sean-k-mooney im not sure i normally get a vote so it was also unilatreral
21:15:26 sean-k-mooney but close enough
21:16:54 efried We don't really vote officially on things like that anyway, so it was all opinion
21:16:58 efried and you know what they say about opinions.
21:17:47 sean-k-mooney :) as an fyi i will be going on vacation in about a week so im going to rebase the two feature im trying to get merged into a chain since they will conflict if i dont
21:20:09 sean-k-mooney for what its worth of the two option i kind of prefer the tradtional filter for filtering nodes by owner
21:21:26 sean-k-mooney if we did it a placemnt way i woudl thing an aggreate feel more natual for grouping RP the a custom trait
21:22:27 sean-k-mooney anyway that is just my gut reaction so i have not really tought about it much
21:27:27 efried sean-k-mooney: we've already established that a trait won't work without hackitude.
21:27:59 sean-k-mooney ack ya noticed the comment form mriedem after i wrote that
21:52:44 efried dansmith: TypeError: Object of type 'datetime' is not JSON serializable
21:52:52 efried This on a cast of prep_resize
21:53:04 efried The payload appears to have datetime.datetime objects in the image meta
21:53:24 efried what's odd is, this wasn't failing before I changed something in a totally different place
21:54:49 dansmith efried: nothing obvious, other than that oslo.serialization has some quiet handling of that
21:55:07 dansmith and there was also some proposed change recently specifically about that, but I think we determined it wasn't going to be a problem
21:57:39 efried that's pretty bizarre
21:57:52 dansmith what is?
21:58:10 efried Is it possible this stack trace is a red herring and the problem is really somewhere else?
21:58:42 dansmith isn't that pretty much always a true statement? :)
22:08:53 melwitt mriedem, efried: lol, you rascals (just read the mentions about request groups)
22:09:19 efried you got ToRed
22:09:23 efried as my father would say
22:09:30 efried (Transfer of Responsibility)
22:14:54 efried dansmith: I think my venv might have just been out of date. Repro'd same result against master. Rebuilding...
22:16:09 efried yuh, that cleared it up.
22:17:51 sean-k-mooney stale pyc files
22:17:56 sean-k-mooney maybe
22:18:15 sean-k-mooney still an odd error
22:18:17 efried no, don't we get bad magic number when that happens? In this case I'm betting it was a downrev of oslo.serialization or similar.
22:18:31 efried don't care enough to track it down exactly.
22:43:42 openstackgerrit Eric Fried proposed openstack/nova master: Use provider mappings from Placement (mostly) https://review.opendev.org/696992
22:44:50 efried mriedem, dansmith, sean-k-mooney, gibi, stephenfin: I think this is done ^
22:44:50 efried It uses placement 1.34 to get rid of the hack where we were calculating mappings based on allocations, in all code paths except revert_resize.
22:45:52 efried and knocks down another two microversions in the parity gap.
22:53:21 sean-k-mooney oh thi si sremoving the code we had on the nova side to compute that and using the info returned form placement correct
22:54:24 efried yeah, unfortunately it doesn't remove the code yet, cause it's still needed by revert_resize. Haven't figure out how we're going to fix that yet. Maybe I leave that to gibi :P
22:54:50 sean-k-mooney will we need to keep it for upgrade too?
22:54:53 sean-k-mooney for n-1
22:55:19 efried I wouldn't have thought so. The data is identical (for existing use cases), just the method of computing that changed.
22:55:34 efried We no longer support downrev placements, you gotta upgrade placement first.
22:56:13 sean-k-mooney sure but did we ever need to look this up on the compute node or was it only used in the schduler
22:56:43 efried um
22:56:55 efried yeah, I think we wanted to be able to use it from the compute
22:56:57 efried but hadn't yet.
22:57:22 sean-k-mooney ok if we havent yet its not an upgrade issue :)
22:57:35 sean-k-mooney we will have the placment native way before we do
22:57:38 efried Like, to correlate VGPUs with the proper mdevs, and eventually to make sure we pin CPUs from the right NUMA nodes.
22:57:50 sean-k-mooney yep that is what i was thinking of
22:58:07 sean-k-mooney we will need to use the mapping when doing the physical assignment on the compute node
22:58:20 sean-k-mooney to make sure we assign the resocue that corralates with the allocation
22:58:22 efried actually, I guess bw gets around that today because the mapping is stuffed into the binding profile, so (I'm guessing) the compute pulls it from there.
22:58:33 efried to... figure out which NIC to assign?
22:58:49 efried start talking network stuff, I hear woh woh woh
22:58:51 sean-k-mooney in the sriov case maybe
22:58:57 sean-k-mooney in the ovs case it does not matter
23:00:04 sean-k-mooney actully in the sriov cas ethe port request has the pf netdev name and we add that to the pci_request object
23:00:17 sean-k-mooney so again stuffing stuff in the vif object to avoid the problem
23:01:23 efried I think ima rebase on the "skip the bw test" patch, and put one on top unskipping to prove it works.
23:01:31 efried so I don't have to recheck a zillion times on that stupid multi-network bug.
23:08:27 openstackgerrit Eric Fried proposed openstack/nova master: Use Placement 1.34 (string suffixes & mappings) https://review.opendev.org/696418
23:08:27 openstackgerrit Eric Fried proposed openstack/nova master: refactor: RequestGroup.is_empty() and .strip_zeros() https://review.opendev.org/696991
23:08:28 openstackgerrit Eric Fried proposed openstack/nova master: Tie requester_id to RequestGroup suffix https://review.opendev.org/696946
23:08:28 openstackgerrit Eric Fried proposed openstack/nova master: Use provider mappings from Placement (mostly) https://review.opendev.org/696992
23:08:29 openstackgerrit Eric Fried proposed openstack/nova master: Restore test_minbw_allocation_placement in nova-next job https://review.opendev.org/697586
23:13:58 efried mriedem, gibi: FYI, re-restoring that minbw test via https://review.opendev.org/#/c/697586/
23:14:30 efried It's at the top of that series for now so I can prove the top patch works, can move it later if necessary.
23:15:27 efried and with that, I'm outta here o/
23:21:59 openstackgerrit sean mooney proposed openstack/nova master: Block rebuild when NUMA topology changed https://review.opendev.org/687957
23:21:59 openstackgerrit sean mooney proposed openstack/nova master: Disable NUMATopologyFilter on rebuild https://review.opendev.org/689861
23:22:00 openstackgerrit sean mooney proposed openstack/nova master: support pci numa affinity policies in flavor and image https://review.opendev.org/674072
23:23:52 sean-k-mooney also rebased that on top of the same skip patch to avoid the rechecks
#openstack-nova - 2019-12-06
00:41:15 mriedem efried: ack yeah i was thinking about pushing a revert on top of the skip that depends on the tempest fix, but you beat me to it
01:22:13 openstackgerrit Merged openstack/nova stable/train: compute: Use long_rpc_timeout in reserve_block_device_name https://review.opendev.org/696953
01:49:06 openstackgerrit Merged openstack/nova stable/train: compute: Take an instance.uuid lock when rebooting https://review.opendev.org/696151
01:49:13 openstackgerrit Merged openstack/nova stable/ocata: Add regression recreate test for bug 1830747 https://review.opendev.org/663143
01:49:13 openstack bug 1830747 in OpenStack Compute (nova) ocata "Error 500 trying to migrate an instance after wrong request_spec" [High,In progress] https://launchpad.net/bugs/1830747 - Assigned to Matt Riedemann (mriedem)

Earlier   Later