Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-02-19
09:37:09 lyarwood Morning all, https://review.opendev.org/#/c/708126/ - any non-RH cores able to review this? Looking to close it out and backport if possible.
09:37:49 lyarwood stephenfin: ah so you're also trying to land on 2.82, is there any etiquette around multiple changes trying to land on the same microversion between M2 and M3?
09:38:02 lyarwood stephenfin: for context https://review.opendev.org/#/c/701430/
09:38:44 stephenfin lyarwood: whoever gets there first? :)
09:38:54 lyarwood stephenfin: ack, may the best change win
09:39:13 stephenfin -2 -2 -2 -2 -2
09:39:22 stephenfin sorted.
09:39:22 lyarwood lol
09:39:29 lyarwood MINE MINE MINE
09:49:27 gibi lyarwood: put https://review.opendev.org/#/c/708126/ into my review queue
09:49:47 lyarwood gibi: awesome thank you
09:58:35 stephenfin bauzas: Want to help me close out my oldest review today? :) https://review.opendev.org/#/c/399653/
09:59:10 gibi lyarwood, stephenfin, rosmaita: I have one question regarding the release notes https://review.opendev.org/#/c/708126/3/releasenotes/notes/absolutely-non-inheritable-image-properties-85f7f304fdc20b61.yaml@19
09:59:12 kashyap stephenfin: Wonder if I can borrow your rST eyes for a quick once-over for a QEMU upstream doc: https://kashyapc.fedorapeople.org/qemu_v4.2.0-1300-g5da20ed7e3_docs/system/qemu-cpu-models.html
09:59:58 kashyap stephenfin: Reading from the bottom, do you find the QEMU / libvirt examples with bullet / sub-bullet combo ugly?
10:00:32 kashyap Or "unclean", or "inelegant", or "un-rST" or ...
10:00:57 stephenfin gibi: Replied. It was done in change https://review.opendev.org/#/c/707738/
10:01:37 stephenfin kashyap: you mean the "Syntax for configuring CPU models" section?
10:01:46 kashyap stephenfin: Exactly
10:02:19 stephenfin Yeah, I tend to refrain from bullet points in favour of definition lists or simply sub-titles
10:02:27 stephenfin though the former wouldn't work here
10:03:21 stephenfin For things like the "Important CPU features for AMD x86 hosts" section though, definitely definition lists for the flags, IMO
10:03:30 kashyap stephenfin: I'll turn them into sub-titles, then.
10:03:41 gibi stephenfin: thanks. I see now. I'm +A
10:03:50 stephenfin kashyap:
10:03:53 stephenfin definition term
10:04:00 stephenfin definition definition
10:04:09 stephenfin no newlines between the two
10:06:00 kashyap stephenfin: I see, okay, you suggest that for the CPU features for AMD|Intel hosts. (I'll ACK your review in the QEMU commit message, of course.)
10:09:25 bauzas stephenfin: oh sorry, missed your ping
10:17:51 bauzas stephenfin: call it a deal
10:22:00 brinzhang A follow up patch of bp/support-delete-on-termination-in-server-attach-volume
10:22:02 brinzhang https://review.opendev.org/#/c/680655/
10:22:38 brinzhang Anyone can review this? It has been pushed for a long time
10:35:50 openstackgerrit Lee Yarwood proposed openstack/nova stable/rocky: Reject boot request for unsupported images https://review.opendev.org/708608
10:38:13 openstackgerrit Lee Yarwood proposed openstack/nova stable/queens: Reject boot request for unsupported images https://review.opendev.org/708609
10:50:52 gibi brinzhang: it would be nice to get gmann and or takashi vote on that followup as that fixes their comments
10:51:40 brinzhang gibi: yes, thanks
10:53:25 brinzhang gibi: and that update gmann and takashi's comments, so I think it would be ok now, and in PS4, they are all +1 for that :)
10:55:39 gibi brinzhang: I see takashi's +1 but I only see -1 from gmann. anyhow I queued that path for review on my side
10:56:22 stephenfin bauzas: Reviewed https://review.opendev.org/#/c/670112/ - just two changes requested and I'm happy
10:56:33 brinzhang gibi: yeah, thanks, I was already fixed gmann's comments, and wait for gmann check for that
10:56:59 brinzhang gmann: please check again https://review.opendev.org/#/c/680655/9
11:03:06 openstackgerrit Stephen Finucane proposed openstack/nova master: Avoid PlacementFixture silently swallowing kwargs https://review.opendev.org/701754
11:05:05 stephenfin bauzas: Got another suppppper easy one here for you https://review.opendev.org/#/c/706466/
11:19:13 openstackgerrit Lee Yarwood proposed openstack/nova stable/stein: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708615
11:19:41 openstackgerrit Lee Yarwood proposed openstack/nova stable/rocky: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708616
11:19:59 openstackgerrit Lee Yarwood proposed openstack/nova stable/queens: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708617
11:20:30 openstackgerrit Lee Yarwood proposed openstack/nova stable/stein: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708615
11:24:56 openstackgerrit Lee Yarwood proposed openstack/nova stable/stein: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708615
11:25:02 lyarwood gah, I need to repost some of these, sorry!
11:25:31 openstackgerrit Lee Yarwood proposed openstack/nova stable/rocky: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708616
11:25:47 openstackgerrit Lee Yarwood proposed openstack/nova stable/queens: Make RBD imagebackend flatten method idempotent https://review.opendev.org/708617
11:25:51 lyarwood okay done, lyarwood--
13:06:29 stephenfin lyarwood: Question for you at https://review.opendev.org/#/c/706897/2/nova/virt/images.py@53
13:32:10 mnaser dansmith: the backfilling issue is confirmed to be resolved, just did it against a cloud with ~450kish records in instances table
13:32:20 mnaser so it didnt try to backfill the hidden value
13:58:30 openstackgerrit Merged openstack/nova master: Absolutely-non-inheritable image properties https://review.opendev.org/708126
13:58:40 openstackgerrit Merged openstack/nova master: tests: Validate huge pages https://review.opendev.org/399653
14:34:13 bauzas stephenfin: ack for the placement audit review and thanks, will look
14:36:29 bauzas stephenfin: and I added a better explanation in https://review.opendev.org/#/c/706466/
14:36:38 bauzas that won't be in the commit msg but okay
14:44:12 dansmith mnaser: sweet, that's been backported and released now so hopefully nobody else will hit it
14:44:18 dansmith mnaser: thanks for being the front line :)
14:49:06 mnaser dansmith: yep, thanks for helping out in nailing down the root cause. I appreciate it.
14:49:16 dansmith np
14:56:27 openstackgerrit Dongcan Ye proposed openstack/nova master: libvirt: Check the guest support UEFI https://review.opendev.org/685678
15:07:50 lyarwood stephenfin: ack thanks, I'll reply shortly
15:20:35 dansmith sean-k-mooney: so looking into my failures from yesterday, I'm actually getting an error from placement:
15:20:52 dansmith sean-k-mooney: "There is more than one numbered request group in the allocation candidate query but the flavor did not specify any group policy. This query would fail in placement due to the missing group policy."
15:20:54 dansmith efried: ^
15:21:38 dansmith my flavor is pretty vanilla with the exception of an accel:device_profile key
15:21:39 efried interesting; there should be gibi code defaulting the group_policy in this case. Perhaps the dp groups are being added after that.
15:22:04 gibi hm hm
15:22:27 dansmith but it sounds like it works (past this point) for sean-k-mooney and was working in the tempest job last I looked
15:22:27 gibi the defaulting code checks what groups are coming from the flavor and what are from the ports
15:24:01 gibi we only default to none if the flavor does not specify groups
15:24:05 efried dansmith: oh, is that warning actually preceding a failure?
15:24:48 dansmith efried: yeah, "got no acs", but I guess I thought it was saying "expect no results shortly"
15:24:49 efried sorry, I'm interleaving. Sounds like this is working as designed, defaulting the group_policy to 'none'.
15:24:59 dansmith ahhh
15:25:10 dansmith I see the real error, it was hidden in the json:
15:25:17 dansmith Invalid resource class in resources parameter: No such resource class CUSTOM_ACCELERATOR_XYZ
15:25:26 dansmith duh sorry
15:25:59 dansmith I'm not sure where that comes from or what I'm supposed to do to make this work in devstack
15:26:00 efried would there be a better way to percolate that error up?
15:26:17 dansmith I would have expected it to fully configure a fake driver, compute, and traits/classes
15:26:18 efried dansmith: I think ordinarily cyborg would create the resource class. In devstack you can create it manually.
15:26:35 efried yeah, good point, the "fake driver" should probably do it.
15:26:51 dansmith I'm not sure where "CUSTOM_ACCELERATOR_XYZ" comes from either
15:27:01 efried That would be in the device profile.
15:27:05 dansmith from cyborg I guess, stating what the resources for the profile should be?
15:27:07 dansmith yeah,
15:27:12 dansmith so ... seems like it should have done more
15:27:20 efried agree
15:27:57 dansmith efried: I was asking sean-k-mooney this yesterday, but will the fake driver expose the inventory for me as well or do I need to hack up some of that?
15:29:01 efried dansmith: I don't know what it *does* do, but I feel like ultimately it *should* do all of that.
15:29:16 gibi for me the cyborg fake driver created FPGA rc and inventory with a CUSTOM_FAKE_DEVICE trait
15:29:19 dansmith yeah, looking for the former
15:29:59 dansmith now I get complaints about two traits that don't exist, so I guess I need to create those too?

Earlier   Later