Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2020-12-03
12:22:08 gibi sean-k-mooney, brinzhang_: what if we merge the changes gradually but not expose the new microversion until the last patch?
12:22:47 gibi what happens if we don't do this change until the last patch? https://review.opendev.org/c/openstack/nova/+/764292/5/nova/api/openstack/api_version_request.py#248
12:23:40 sean-k-mooney gibi: i was wondering about that too i think we did that in the past but if we do we need to diable the new codepaths until the last patch merges
12:24:59 brinzhang_ for example: in the server list apis, I need to remove the tenant_id and all_tenants, and replaced with project_id and all_projects, some places need the version to have a judgement, maybe it's not easy to control all changes if there is not a microversion
12:25:00 gibi sean-k-mooney: why do we need to disable the new codepaths? Not having the microversion 88 published makes those new codepath unreachable
12:25:12 brinzhang_ it's my opinion
12:25:22 sean-k-mooney gibi: oh ok in that case yep that should work
12:25:33 gibi brinzhang_: you do have a microversion it just not available for the client to request it
12:25:39 sean-k-mooney unless a different api change takes the microversion
12:26:00 gibi sean-k-mooney: yeah, another api change can break this
12:26:05 gibi hm hm
12:26:56 gibi so if we want to go this direction then we need to make the api patches serialized
12:27:03 gibi right now
12:27:09 gibi not just at the merge point
12:27:23 sean-k-mooney basically allocate a micro version to them
12:27:44 sean-k-mooney could we do that with placeholder microverions?
12:28:05 sean-k-mooney and change the meaning of the microversion when it merges
12:28:10 sean-k-mooney that a bit of hack
12:28:19 sean-k-mooney bug kind of like the palceholder db migrations
12:28:53 gibi hm
12:29:00 sean-k-mooney allocate 2.88 2.89 and 2.90 to the 3 specs in flight and make them all the same as 2.87 for now
12:29:09 brinzhang_ as your concern above, if we placeholder microversion, and I am not completed this in this release, it will be a bad news
12:29:13 sean-k-mooney that still not great
12:29:42 gibi how far the other two patches from merge?
12:29:56 gibi if we could merge them quickly then this problem is solved for the cycle
12:30:06 gibi (assuming no new spec proposed api changes for wallaby)
12:30:36 gibi e.g. we allocating the microversion by merging the two small feature first
12:30:40 brinzhang_ there is a noVNC spec reqired a microversion also in my list
12:31:46 gibi so the novnc password, and the hypervisor api modernization are the in flight api changes besides the tenat_id one
12:32:45 brinzhang_ yes, if there are all quickly completed fastly, it would be better.
12:33:24 gibi I don't see open patches for the hyperisor modernizaton :/
12:33:50 brinzhang_ As I know the noVNC implentation code is ready to review, my staff complete it today, all tests: https://review.opendev.org/c/openstack/nova/+/622336
12:34:42 sean-k-mooney well there is also https://review.opendev.org/c/openstack/nova/+/749505
12:35:18 sean-k-mooney that one proably shoudl be merged first for 2.88
12:35:31 sean-k-mooney although it could be after
12:35:54 sean-k-mooney we do wnat to get that done sooner rather then later however
12:36:47 sean-k-mooney stephenfin: whats your plance regardign fixing all the http status codes ^
12:37:11 stephenfin I don't really have one. It sounds very low priority. I might do it if I get some other stuff landed first
12:37:18 sean-k-mooney stephenfin: looks like you havent worked on it in 3 months will you be updating that soon or in the new year
12:37:43 sean-k-mooney ok so sound like we shoudl not wait for that one so
12:37:53 stephenfin no, definitely not
12:38:50 gibi stephenfin: thanks for the info
12:39:09 stephenfin do I note have the hypervisor one up?
12:39:12 stephenfin not
12:39:19 sean-k-mooney you do i think
12:39:30 zigo Hi. When building Nova 22.0.1, I get 1787 unit test failures like this one: http://paste.openstack.org/show/800690/
12:39:30 zigo while nova 22.0.0 does build without any issue.
12:39:30 zigo Is this a known problem?
12:39:34 gibi I don't find it here https://review.opendev.org/q/topic:%22bp%252Fmodernize-os-hypervisors-api%22+(status:open%20OR%20status:merged)
12:39:59 stephenfin whoops, wrong topic
12:39:59 stephenfin https://review.opendev.org/q/topic:%22bp%252Fdeprecate-os-hypervisors%22+(status:open%20OR%20status:merged)
12:40:28 brinzhang_ so the noVNC that we can promote priority, maybe it's the fast one that closes to merge
12:40:59 stephenfin gibi: fixed
12:40:59 sean-k-mooney brinzhang_: gibi have we been doing runway this cycle
12:41:10 zigo If nobody knows, then I can probably bisect the 22.0.0 to 22.0.1 patches ...
12:41:25 gibi zigo: give me a se
12:41:25 gibi c
12:41:34 gibi stephenfin: thanks
12:41:36 sean-k-mooney not really
12:41:49 sean-k-mooney ok can we add the password one to https://etherpad.opendev.org/p/nova-runways-wallaby
12:42:01 gibi sean-k-mooney: the etherpad is up there it is just only brinzhang_ who put thing is nit
12:42:06 gibi thing in it
12:42:19 brinzhang_ sean-k-mooney: yeah, and I already used it to put cyborg support shelve/unshelve feature, and I will add the noVNC patch later in the runway
12:44:42 gibi zigo: looked at the diff but nothing popped out, so I guess it is bisect time.
12:45:02 zigo gibi: Thanks for looking, I'll try then.
12:45:12 gibi zigo: let me know what you found
12:45:30 sean-k-mooney yep so if we do noVNC then os-hyperviors then teants change?
12:46:10 sean-k-mooney that shoudl avoid it assumeing stephens paatch is already ready for review otherwise move os-hyperviors to 2.90
12:48:08 gibi sean-k-mooney: as soon as we merging the first patch of the tenant_id referring internally to 2.88 (or any specific microversion) we commit to not have any other microversion bump merged until we finish the tenant_id work. I'm affraid of such commitment
12:49:02 brinzhang_ gibi: agree
12:49:58 brinzhang_ if I cannot complete it in Wallaby release, the 2.88 just suit for part of apis
12:51:05 brinzhang_ but if wait for all patches submited, and ready to review, there will be a heavy work
12:52:13 gibi there is no easy solution
12:52:14 sean-k-mooney brinzhang_: merging it would block all other api work however so really we need ti to be close before we merge the first patch
12:52:25 vinay_m hello every one, i have a devstack setup on a baremetal, and i have created a image out of snapshot of a openstack VM install with vepc software,when im launching vms with this image not able to ping or ssh , while vms with cirros or some other works fine
12:53:05 vinay_m not able to figure out what could be issue
12:53:16 gibi brinzhang_: let's make it so that I will review the open API patches in tenant_id, vnc and hypervisor series to see where we are now.
12:53:44 sean-k-mooney vinay_m: did you look at the console output to see if cloud init reran and updated the networking correctly
12:53:47 zigo It's super annoying: invidual unit tests are looking good, it's only when I start testing them all at once that I get failures ...:/
12:54:31 gibi zigo: sh*t, I would like to help you. Is there any specific env setup that makes this appear? as I don't see it happening in the gate
12:54:39 vinay_m sean-k-mooney: yeah tried , but no log from this vm
12:54:42 sean-k-mooney vinay_m: if you had hardcoded macs or static ips the vm might just nto be getting the right ip
12:54:52 zigo gibi: I'll bisect and find out.
12:54:55 gibi zigo: anyhow I started running unit test on 22.0.1 in a loop now locally
12:55:02 zigo But it will take a long time...
12:55:09 sean-k-mooney vinay_m: there shoudl always be at least a minimal log is it actully booting correct
12:55:31 brinzhang_ gibi: the noVNC patch is ready to review now, IMO, it can be mereged firstly
12:55:43 vinay_m sean-k-mooney: same doubt , but status was showing active that's it
12:56:46 brinzhang_ the tenant_id just ready for the list servers apis change, except the policies changes, but after I get gmann's sugestions, I think it's easily to ready
12:57:18 sean-k-mooney vinay_m: and nothing form "openstack console log show <uuid>"?
12:57:39 brinzhang_ gibi: I am not similiar with the hypervisor series process
12:57:51 vinay_m sean-k-mooney: no console log
12:58:04 sean-k-mooney vinay_m: you might need to connect to libvirt directly and check but it sound like its not even getting to finding the os
12:58:14 gibi brinzhang_: I think you have to make progress with the tenant_id patches while we merges the other API impacts and then try to assess the situation to see if your patches can be merged in once
12:59:00 sean-k-mooney vinay_m: im not familar with vepc but if it used differnt disk buses or machine types it could cause the image to be unbootable
12:59:35 gibi sorry but from the overall nova developement progress perspective I have to prioritize the smaller API impacts first before we block the API changes due to the tenant_id work.
12:59:37 vinay_m sean-k-mooney: os i have used was ubuntu 18.0.4
12:59:39 sean-k-mooney oh vepc is vEPC
13:00:03 vinay_m <sean-k-mooney> oh vepc is vEPC -> yes :-)
13:00:25 brinzhang_ From my current job, I may not be able to complete all of them, because I have to update the Cyborg Shelve /unshelve support patch, or the noVNC patch with my partner, and other jobs from my boss

Earlier   Later