Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2024-09-17
16:20:33 bauzas #topic vmwareapi 3rd-party CI efforts Highlights
16:20:38 bauzas fwiesel: anything to report ?
16:20:43 fwiesel #info No updates
16:21:00 fwiesel Not really... patch still stuck in ci... I'll recheck it soon
16:21:13 bauzas cool
16:21:17 bauzas last topic then
16:21:26 bauzas #topic Open discussion
16:21:33 bauzas (noonedeadpunk) Enable serving NoVNC under subpath (not only via port/subdomain)
16:21:38 bauzas noonedeadpunk: around ?
16:21:48 bauzas #link https://review.opendev.org/c/openstack/nova/+/928839
16:22:50 sean-k-mooney so its tecnnically possible ot do ^ externally with the correct revers proxy config but its a bit of a pain
16:23:11 sean-k-mooney im generally supportive of this as a minor feature (specless blueprint)
16:23:16 sean-k-mooney tl;dr
16:23:40 sean-k-mooney if you want to deploy all services on port 443 including novnc
16:23:49 gibi I only skimmed it, is this only support a single new query param "path"?
16:24:05 sean-k-mooney (either beacuse of yoru isp or firewall issues) you can do it with subdomains but its tricky to do with subfolders
16:24:25 sean-k-mooney yep path is the subpath to append ot the websocket
16:24:43 sean-k-mooney so in there case they wanted to deply openstac.my.domain/vnc
16:24:50 bauzas I'm having a few concerns by the implementation but I don't have any concenr about the design
16:25:01 sean-k-mooney without setting path it will connect ot /
16:25:02 noonedeadpunk yeah, around jsut a bit busy
16:25:27 bauzas like, I don't want nova to just sub the path and see whether there is path
16:25:33 sean-k-mooney so this is just exposing a way to pass path to novnc
16:25:44 bauzas maybe a config option, maybe something like a relnote
16:26:08 noonedeadpunk nova does not sub the path - but it jsut respects whatever user defiens in config
16:26:09 bauzas but I don't want us to not expose that
16:26:25 noonedeadpunk by not respecting query, and not just fqdn
16:26:33 bauzas well your change will parse the qs
16:26:43 noonedeadpunk and append to it
16:27:09 sean-k-mooney so we brifly dicussed having a config option like the exisitng port option for the query sting
16:27:24 noonedeadpunk vs currnet behaviour where query is hardcoded basically
16:28:08 sean-k-mooney for libvirt we decieded that just passing a uri for the connection was better
16:28:17 noonedeadpunk well, I'm not sure about config option ,as there could be more usecases whre you wnat to pass smth as admin to novnc
16:28:17 bauzas well, here you're hard parsing the query with only 'path'
16:28:24 sean-k-mooney so that why i was mostly ok with noonedeadpunk approch as its consitent with that
16:28:54 sean-k-mooney bauzas: there may be other query arges that novnc uses tha tcould be reasoanable to pass
16:29:02 noonedeadpunk bauzas: yeah, but current code will jsut add `?path=?token=<uuid>`
16:29:04 sean-k-mooney so i woudl prefer not to limit it to path
16:29:37 bauzas anyway, as I said, this is a technical detail
16:29:43 noonedeadpunk so if you try to add like anything to the `novncproxy_base_url` - you will end up with invalid token
16:29:54 bauzas I don't oppose to the usecase and I think keeping an URI is important
16:30:13 noonedeadpunk probably I jsut don't get what you're proposing isntead...
16:30:35 bauzas as I said, I only have implementation concerns
16:30:51 bauzas your current patch only accepts 'path' as a qs param
16:30:59 noonedeadpunk but like unit tests shows it should be quite consistent... I can add more test cases if I'll get what these concerns are
16:31:35 bauzas if we restrict to 'path' then we need to document it better than "oh and now you can pass *any* qs"
16:31:42 sean-k-mooney noonedeadpunk: bauzas is askign to make sure it will work with any arg
16:31:51 noonedeadpunk it's not restricted with path?
16:32:14 noonedeadpunk l133 of https://review.opendev.org/c/openstack/nova/+/928839/3/nova/tests/unit/objects/test_console_auth_token.py
16:32:21 bauzas "@item(hotlist):button2
16:32:24 bauzas doh
16:32:46 bauzas the documentation says "You can also supply extra request arguments"
16:33:25 bauzas anyway, can we just (for the sake of my beer) accept that blueprint as specless and nitpick over the change as much as we want over the Epoxy timeframe ? :)
16:33:34 noonedeadpunk ++
16:33:43 sean-k-mooney +1 form me also
16:34:04 bauzas any objection ?
16:34:48 bauzas looks not
16:35:10 bauzas #agreed https://blueprints.launchpad.net/nova/+spec/novnc-base-url-respect-extra-params accepted as specless for the Epoxy release
16:35:17 bauzas I'll file the paperwork
16:35:22 gibi OK for me
16:35:24 bauzas that's it for the agenda
16:35:39 bauzas any other bit you'd like to raise ?
16:35:56 bauzas looks not
16:36:06 bauzas my beer won't heat, cool
16:36:10 bauzas thanks all
16:36:13 bauzas #endmeeting
16:36:14 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2024/nova.2024-09-17-16.01.log.html
16:36:14 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2024/nova.2024-09-17-16.01.txt
16:36:14 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2024/nova.2024-09-17-16.01.html
16:36:14 opendevmeet Meeting ended Tue Sep 17 16:36:13 2024 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
16:36:17 sean-k-mooney o/
16:36:22 elodilles thanks o/
16:38:23 gibi o/
#openstack-nova - 2024-09-18
01:14:48 opendevreview Masahito Muroi proposed openstack/nova master: Use boot_roles_count and boot_role_ key for system_metadata https://review.opendev.org/c/openstack/nova/+/925163
04:43:48 opendevreview OpenStack Proposal Bot proposed openstack/nova master: Imported Translations from Zanata https://review.opendev.org/c/openstack/nova/+/929719
06:08:52 gibi bauzas: it seems to me we are still blocked https://review.opendev.org/c/openstack/nova/+/928310/comments/b3c9bcb6_9be616a6
06:11:45 bauzas eek ok I see
06:13:46 bauzas https://github.com/openstack/python-openstackclient/blob/master/openstackclient/compute/v2/server.py#L3875
06:14:15 bauzas I'm lost now
06:14:26 bauzas we still send admin_password to evacuate
06:14:39 bauzas oh, but that's to the sdk
06:17:43 bauzas and I don't see 7.1.1 here https://docs.openstack.org/releasenotes/python-openstackclient/unreleased.html
06:18:17 bauzas elodilles: ^
06:21:39 bauzas https://github.com/openstack/python-openstackclient/compare/7.1.0...7.1.1
06:22:12 bauzas oh fuck, I understand it
06:22:54 bauzas https://github.com/openstack/openstacksdk/blob/master/openstack/compute/v2/server.py#L748
06:23:15 bauzas this is neither 'password' nor 'admin_password', but just 'admin_pass'
06:23:27 bauzas we lost 3 days basically and we need another fix and another release
06:23:46 bauzas stephenfin: elodilles: frickler: ^
06:47:57 bauzas gibi: https://review.opendev.org/c/openstack/python-openstackclient/+/929726
06:48:18 bauzas this is 8.47am and I already have headache
06:50:29 bauzas frickler: elodilles: I'm in favor of exceptionnally rollbacking to OSC6.x for Dalmatian, we are always awfully late for Nova RC1
07:00:48 gibi bauzas: quickly reproduced the osc issue in a fresh devstack, then applied your patch and now I have proof that it works
07:01:02 gibi thanks for quickly proposing the fix
07:30:10 gibi bauzas: if you are interested in the microversion 2.95 evacuate issue I mentioned in my comment in the osc patch then https://bugs.launchpad.net/nova/+bug/2081023 This is probably just a cosmetic issue.
07:47:53 frickler thx, I commented on this yesterday already in the qa channel, I was under the impression that sean-k-mooney had tested my change, sorry for that. is this option also really named differently between this and what create and rebuild use? might be just an sdk thing, but certainly not helpful
07:48:25 frickler I'm also not really here this week, so try to ping other sdk, release and reqs people as needed
08:12:50 bauzas sorry folks, I was off the laptop last hour
08:12:52 bauzas now I'm back
08:14:39 bauzas so, I'd prefer to rollback the u-c by OSC6 now

Earlier   Later