Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-10-08
15:06:30 openstackgerrit Stephen Finucane proposed openstack/nova master: docs: Add image metadata property reference guide https://review.opendev.org/756867
15:27:06 stephenfin bauzas: tox config isn't a API we need to protect :) We change things in there (and rightly so) all the time https://review.opendev.org/#/c/756135/
15:27:37 bauzas stephenfin: sure, but I want to make sure that scripts still work and be a gentleman
15:27:49 bauzas I'm currently looking at how we could do it
15:27:52 stephenfin the only scripts using that are in zuul, which uses both
15:27:58 stephenfin but prefers the new style
15:28:30 bauzas honestly, I'm not that opiniated
15:28:55 bauzas stephenfin: are we sure that all CI (meaning 3rd party) won't use it too ?
15:30:10 stephenfin bauzas: I very much doubt it. The only reason to do so is if you wanted to use a different set of upper constraints than master. Also, they've probably hit it elsewhere http://codesearch.openstack.org/?q=TOX_CONSTRAINTS_FILE&i=nope&files=&repos=
15:31:03 stephenfin *different set of upper constraints than upstream/opendev.org
15:32:25 bauzas stephenfin: well, surely, but we can't assure we won't break others
15:32:41 stephenfin bauzas: actually, we can
15:32:41 bauzas stephenfin: that's the reason why I do care to be soft :)
15:32:55 stephenfin look at the zuul checks
15:32:59 stephenfin no failing third party CIs
15:33:17 stephenfin (in fact, no third-party CI results whatsoever)
15:34:12 sean-k-mooney if its overly complicated im fine with just changing it i like haveing the url in only one place however in the file
15:34:21 sean-k-mooney that the main thing i didnt like
15:34:37 sean-k-mooney we we can support both and prefer teh new form then that fine too
15:34:42 bauzas okay you know what ?
15:34:50 sean-k-mooney but if its overly complex lets just keep it simple
15:35:03 bauzas this change is taking too much of my time, I'll just quit the discussion with a +W
15:35:17 bauzas let's over bikeshed it, it's not worth it
15:35:25 bauzas not* bikeshed
15:37:34 sean-k-mooney sure i +1ed becasue i was not going to block over it i just did not want to merge it unless other agreeed as i prefered it teh way it was before
15:46:03 gibi nova meeting starts in 15 minutes on #openstack-meeting-3
16:11:15 stephenfin kashyap: You'd probably be interested in this https://bugs.launchpad.net/nova/+bug/1898715
16:11:15 openstack Launchpad bug 1898715 in OpenStack Compute (nova) "Live migration fails despite matching CPUs" [Undecided,New]
16:11:17 stephenfin Seems an easy fix
16:25:28 kashyap stephenfin: Will look; but from a quick first look, great sleuthing there by the reporter
16:25:51 kashyap The CPU model stuff...never ending joy
16:54:53 openstackgerrit Balazs Gibizer proposed openstack/nova master: Move query param generation to RequestGroup o.vo https://review.opendev.org/756894
16:54:59 kashyap stephenfin: Responded to Andrew there on the bug, and also added a note about what on earth does "require" supposed to mean. Thanks for bringing it to my attention :-)
16:55:19 kashyap (The "require" being an libvirt XML bit)
16:56:17 openstackgerrit Balazs Gibizer proposed openstack/nova master: Refactor _claim_pci_device_for_interface_attach to prepare for qos https://review.opendev.org/756895
16:57:38 openstackgerrit Balazs Gibizer proposed openstack/nova master: Support interface attach with qos ports https://review.opendev.org/756530
17:05:47 openstackgerrit Merged openstack/os-vif master: tox: Rename 'UPPER_CONSTRAINTS_FILE' -> 'TOX_CONSTRAINTS_FILE' https://review.opendev.org/756135
#openstack-nova - 2020-10-09
08:26:49 openstackgerrit Balazs Gibizer proposed openstack/nova master: Ignore PCI devices with 32bit domain https://review.opendev.org/756697
09:06:32 lyarwood stephenfin: image defined bdms
09:06:46 lyarwood stephenfin: block_device_mappings in the image metadata or something else?
09:06:55 lyarwood stephenfin: oh img_mappings
09:09:09 lyarwood stephenfin: I didn't think we used it anymore tbh
09:09:21 lyarwood stephenfin: I'll take a better look later today
09:41:14 openstackgerrit wangzhh proposed openstack/nova master: Ignore InstanceNotFound exception when host rejoin cluster https://review.opendev.org/757053
10:21:12 sean-k-mooney lyarwood: we use it here https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L833-L856
10:23:44 sean-k-mooney which is used here https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L907
10:25:33 lyarwood sean-k-mooney: right we use block_device_mapping, I don't think we use img_mappings anywhere
10:25:42 sean-k-mooney we do
10:25:52 sean-k-mooney https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L933-L934
10:26:04 sean-k-mooney we merge the image mappings with the block device mappings
10:26:21 sean-k-mooney before we append the flavor mappings
10:26:48 sean-k-mooney so right now we get the mappigns form the commandline the falvaor and iamge and merge them all together
10:27:06 lyarwood right but I still don't see any code looking up img_mappings
10:27:22 sean-k-mooney this does it https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L907-L908
10:27:59 sean-k-mooney https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L841
10:28:27 lyarwood sean-k-mooney: mappings != img_mappings
10:28:31 sean-k-mooney in the ovo we store "img_mappings" in the "mappings" field
10:28:34 lyarwood sean-k-mooney: that's my point
10:28:35 sean-k-mooney lyarwood:yes it is
10:29:22 sean-k-mooney https://github.com/openstack/nova/blob/f24e054f5e41dcf45b7ecd20e6d41f4a293a91d2/nova/objects/image_meta.py#L555
10:29:40 sean-k-mooney img_mappings is the legacy name for mappings
10:29:54 sean-k-mooney we coalese it and store it in mappings
10:30:19 lyarwood mappings is the legacy name
10:30:21 sean-k-mooney actully i guess its the other way around
10:30:41 sean-k-mooney ok well the lookup should still work
10:30:59 sean-k-mooney we are using mappings which will retrun the same valuse as img_mappings
10:31:21 sean-k-mooney because of https://github.com/openstack/nova/blob/f24e054f5e41dcf45b7ecd20e6d41f4a293a91d2/nova/objects/image_meta.py#L565-L572
10:31:22 lyarwood I guess the lookup is defined elsewhere?
10:31:56 sean-k-mooney actully that the set where is the get
10:32:03 lyarwood yeah
10:32:32 sean-k-mooney oh we just have it for set so when we build it we set it with the new name
10:33:03 sean-k-mooney the get is https://github.com/openstack/nova/blob/f24e054f5e41dcf45b7ecd20e6d41f4a293a91d2/nova/objects/image_meta.py#L657-L685
10:33:43 lyarwood but https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L841 is specifically using the older name so that's always None
10:34:36 sean-k-mooney so is if from here https://github.com/openstack/nova/blob/9fa563666e8500510def19269f740d99cab0e994/nova/compute/api.py#L843-L845
10:37:07 sean-k-mooney not looking at the img_mappings would be a bug unless we had an explict deprecation of this fucntionaliy
10:38:46 lyarwood I'm still not entirely sure what it is the dict provides tbh, looking for an example
10:39:02 lyarwood virtual image names to device names
10:40:13 sean-k-mooney i think you could pass anything you can pass in bdm on the commandline
10:41:18 lyarwood no that's stored in block_device_mapping or bdm_v2
10:41:42 lyarwood this is simple dict of things like {'root': '/dev/vda'} AFAICT
10:42:17 lyarwood https://github.com/openstack/nova/blob/4cf72ea6bfc58d33da894f248184c08c36055884/nova/block_device.py#L538-L588
10:42:45 sean-k-mooney i see in that case im less concerned
10:43:00 sean-k-mooney we dont provide stable/usersetable device names really anyway
10:43:59 lyarwood yup I wonder what this is actually used for tbh
10:44:41 sean-k-mooney https://github.com/openstack/nova/blob/f24e054f5e41dcf45b7ecd20e6d41f4a293a91d2/nova/objects/image_meta.py#L454-L460
10:46:31 sean-k-mooney it looks like it was never tested when we did the conversion
10:46:36 sean-k-mooney to ovos
10:47:15 lyarwood yeah we should get everything we need from img_block_device_mapping / block_device_mapping
10:48:35 openstackgerrit Lee Yarwood proposed openstack/nova master: compute: Lock by instance.uuid lock during swap_volume https://review.opendev.org/754695
11:24:19 openstackgerrit Lee Yarwood proposed openstack/nova master: Include removal of ephemeral backing files in the image cache manager https://review.opendev.org/689422
11:58:43 openstackgerrit Lee Yarwood proposed openstack/nova stable/ussuri: Use absolute path during qemu img rebase https://review.opendev.org/757083
11:59:07 openstackgerrit Lee Yarwood proposed openstack/nova stable/train: Use absolute path during qemu img rebase https://review.opendev.org/757084
11:59:28 openstackgerrit Lee Yarwood proposed openstack/nova stable/stein: Use absolute path during qemu img rebase https://review.opendev.org/757085
12:48:00 openstackgerrit Stephen Finucane proposed openstack/nova master: docs: Add image metadata property reference guide https://review.opendev.org/756867
12:49:01 stephenfin lyarwood: Context on that img_mappings question was ^ I was trying to document what it was for and remove it if it wasn't necessary. I resorted to marking it as internal only and kicking the can down the road :-)
12:49:21 stephenfin though it sounds like you and sean-k-mooney might have teased it out, /me reads back up
12:50:08 lyarwood stephenfin: yeah tbh I don't think it's being used given what we were talking about above
12:50:10 lyarwood brb
12:51:31 stephenfin sean-k-mooney: I replied to your questions on that doc review. tl;dr: auto generating this stuff is too hard, and I'm not looking to replace Glance's XML schemas for code-based verification purposes
12:53:37 sean-k-mooney stephenfin:ya looking at the code we still have code to use the legacy name

Earlier   Later