| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-01-24 | |||
| 15:31:15 | amrith | in general, I'd think that is something that for trove, should be trove's to monitor | |
| 15:31:19 | mriedem | amrith: have you talked to the requirements team about getting a trove unit test job to run as part of changes like this? https://review.openstack.org/#/c/414170/ | |
| 15:31:32 | amrith | have not | |
| 15:31:34 | andreykurilin | mriedem: because all "right version" from min to max should be compatible. we have too much changes now to use 2.29 as a min (like it was previously) | |
| 15:31:39 | amrith | but it is a good idea as well | |
| 15:32:14 | andreykurilin | I mean it is a hard task to write a compatible code for 2.29 and 7.0 | |
| 15:32:28 | mriedem | andreykurilin: well yeah, and that's why we use semver for major changes | |
| 15:32:49 | mriedem | even with fixing the token thing, 7.0 probably breaks a bunch of things | |
| 15:33:26 | andreykurilin | mriedem: it breaks two things - token auth via python API (for CLI it was broken for a year) and list_extension | |
| 15:33:41 | andreykurilin | both of these things have fixes on review | |
| 15:34:00 | bauzas | 7.0.0 is breaking more than the token thing ? | |
| 15:34:00 | mriedem | well by 'break' i mean, intentional changes to remove functionality | |
| 15:34:12 | mriedem | which is why we released it with a major version bump | |
| 15:34:16 | bauzas | do we have a g-r change that blocks it ? | |
| 15:34:32 | mriedem | bauzas: the requirements changes aren't run against trove unit tests, | |
| 15:34:38 | mriedem | otherwise something might have been caught | |
| 15:34:45 | bauzas | and if we have a couple of fixes, let's name the next release 7.0.1 even if we add the osprofiler thing | |
| 15:34:47 | andreykurilin | mriedem: I think we done everything in backward compatible way (except usage of our inner variables) | |
| 15:35:23 | bauzas | mriedem: yeah, but as per andreykurilin 7.0.0 blew up more than just trove | |
| 15:35:38 | bauzas | mriedem: at least the tripleo folks and the murano folks AFAICS | |
| 15:35:39 | mriedem | bauzas: new features need to be a min version bump at least to 7.1.0 | |
| 15:35:57 | bauzas | I know | |
| 15:35:58 | andreykurilin | bauzas: one more thing that is broken by 7.0 - https://review.openstack.org/#/c/424570/ :) | |
| 15:36:22 | mriedem | on https://review.openstack.org/#/c/423808/ i would have made the min 6.0, and u-c 7.0 | |
| 15:36:22 | bauzas | I'm saying 7.0.0 is doomed so we would signal that there is a new release fixing it | |
| 15:36:57 | andreykurilin | ok, let's use 6.0 as a min, but not 2.29 like it was | |
| 15:36:58 | bauzas | yup :( | |
| 15:37:12 | mriedem | andreykurilin: do you want me to push that change quick? | |
| 15:37:24 | mriedem | >=6.0.0,!=7.0.0 | |
| 15:37:41 | andreykurilin | mriedem: no, please take a rest. you are on vacation | |
| 15:37:47 | andreykurilin | I'll do it | |
| 15:38:10 | mriedem | it's morning and i've got a caffiene headache but ok | |
| 15:38:29 | mriedem | *caffeine | |
| 15:38:36 | bauzas | mriedem: you okay if I'm tagging a 7.1.0 with the help of andreykurilin ? | |
| 15:38:50 | bauzas | once all the fixes are done of course | |
| 15:38:55 | mriedem | bauzas: that's fine | |
| 15:39:19 | mriedem | alex_xu: cdent: artom: gmann: where are we on the device tags thing? | |
| 15:39:20 | bauzas | mriedem: client libraries freeze is thursday as well ? | |
| 15:39:27 | mriedem | bauzas: yes client release freeze is thursday | |
| 15:39:34 | bauzas | ok | |
| 15:40:06 | cdent | mriedem: I don't know, haven't revisited since I lit the fire... | |
| 15:40:27 | mriedem | it's a time that i wish we could just fix the bugs in the api w/o the microversion | |
| 15:40:33 | artom | mriedem, nowhere as far as I can tell - what's the plan? We document before ocata FF and unfuck it in Pike? | |
| 15:41:03 | artom | If so I can submit the documentation patch | |
| 15:41:05 | mriedem | (a) this version works, but (b) x-y don't work, except z, so (c) we'll fix it all in microversion 2.50 | |
| 15:41:11 | artom | But I wasn't sure we had consensus on that | |
| 15:41:15 | robcresswell | diana_clarke: Thanks for the response on the horizon discussion, that was useful | |
| 15:41:49 | mriedem | artom: if we're not going to fix it with a microversion in ocata we can document it probably whenever before ocata rc1 | |
| 15:42:23 | artom | mriedem, do we have time to squeeze it in in 2 days? I can write the code... | |
| 15:42:28 | artom | (The new microversion I mean) | |
| 15:42:52 | mriedem | artom: we probably don't have the review bandwidth in 2 days | |
| 15:43:16 | mriedem | alternatively we microversion it for a bug fix and it's not tied to FF since it's not a feature, it's a bug fix | |
| 15:43:23 | diana_clarke | robcresswell: np, I know the whole approach is somewhat sub-optimal, but I'm not sure how else we could have proceeded. | |
| 15:43:28 | mriedem | i thought we did that with the vm status microversion in newton | |
| 15:43:41 | artom | mriedem, oh right, I forgot FF != halt all evarything | |
| 15:44:33 | artom | mriedem, if I have your PTLness backing me on this, I can drive it as a microversion bugfix before ocata release | |
| 15:44:36 | artom | ;) | |
| 15:44:46 | cdent | I think bugfix sounds like the right approach | |
| 15:44:52 | cdent | gives some time to do it right | |
| 15:44:56 | cdent | and it _is_ a bugfix | |
| 15:45:30 | robcresswell | diana_clarke: Yeah, understood | |
| 15:45:38 | mriedem | especially this one https://review.openstack.org/#/c/304510/64/nova/api/openstack/compute/block_device_mapping.py i'd really like to just fix that w/o a microversion, | |
| 15:45:49 | mriedem | but that probably goes against everything we've always said | |
| 15:47:06 | artom | mriedem, I don't think we can. For all intents and purposes, there's no tag in >=2.33. And adding an API parameter is a new microversion... | |
| 15:48:05 | mriedem | yes i know, but 2.33->2.43 or whatever with no bdm tag support, and then it's back in, is just super terrible UX | |
| 15:48:11 | cdent | bauzas: are your placement changes in a good state to review now? | |
| 15:48:18 | bauzas | cdent: it is indeed | |
| 15:48:23 | cdent | on it | |
| 15:48:45 | bauzas | cdent: I need to write a grenade change for fixing the related jobs but that shouldn't be impacting my WIP | |
| 15:48:46 | artom | mriedem, I totally agree on that point... And to be fair, I can't think of any core that would stop working | |
| 15:49:03 | bauzas | cdent: FWIW, there is now only one patch left, I abandoned the fixture one | |
| 15:49:10 | cdent | k | |
| 15:49:16 | artom | Tags were optional, so if you tried to use them in 2.33 it would break, but no one filed a but so no one noticed | |
| 15:49:32 | artom | But if you weren't using them re-adding them as an optional parameter doesn't break anything | |
| 15:49:43 | artom | Hrmm, maybe we *can* just silently re-add them | |
| 15:49:53 | artom | Same logic goes for networks | |
| 15:50:33 | bauzas | dansmith: quick question on grenade, do you think it's a valid approach to just source the placement lib, and call out the appropriate functions, or try to modify local.conf to include placement? | |
| 15:50:34 | mriedem | i doubt anyone noticed because (a) most clouds aren't at newton yet and/or (b) no one is using them, yeah | |
| 15:50:47 | dansmith | bauzas: source and use existing functions | |
| 15:50:58 | mriedem | however, | |
| 15:50:59 | bauzas | dansmith: that would mean local.conf wouldn't be accurate then | |
| 15:51:25 | mriedem | 2.1 and microversions is all about correcting mistakes, and in the old 2.0 API we always said we couldn't fix things regardless of the wart | |
| 15:51:29 | bauzas | dansmith: if that's not a necessary requirement, I'm fine | |
| 15:51:33 | mriedem | unless we used extensions or some garbage like that | |
| 15:51:38 | dansmith | bauzas: oh, thought you meant nova.conf.. I still think that's okay and consistent, FWIW | |
| 15:52:13 | bauzas | dansmith: the fact is that the grenade-upgrade local.conf wouldn't be valid in case the deployer wants to unstack/stack | |
| 15:52:20 | mriedem | andreykurilin: ping me when your requirements change is up to change the minimum back to 6.0 and blacklist 7.0 | |
| 15:52:27 | cdent | mriedem: sounds like you should join the thread about api compatbility/stability seeing as you have opinions ;)/2 | |
| 15:52:45 | andreykurilin | mridem: it is here https://review.openstack.org/#/c/424705/ | |
| 15:53:14 | mriedem | cdent: well i lean toward the microversion fix | |
| 15:53:24 | mriedem | just because it's the "right thing to do" | |
| 15:53:34 | openstackgerrit | Balazs Gibizer proposed openstack/nova: handle uninited fields in notification payload https://review.openstack.org/415857 | |
| 15:54:18 | openstackgerrit | sahid proposed openstack/nova: numa: add numa constraints for emulator threads policy https://review.openstack.org/385023 | |
| 15:54:21 | openstackgerrit | sahid proposed openstack/nova: objects: add new field cpuset_reserved in NUMACell https://review.openstack.org/400874 | |
| 15:54:21 | openstackgerrit | sahid proposed openstack/nova: libvirt: return a CPU overhead if isolate emulator threads requested https://review.openstack.org/385364 | |
| 15:54:22 | artom | mriedem, it *is* more transparent | |
| 15:54:22 | openstackgerrit | sahid proposed openstack/nova: numa: fit instance topology with cpus reserved onto cores https://review.openstack.org/400875 | |
| 15:54:23 | openstackgerrit | sahid proposed openstack/nova: virt: update overhead to take into account vCPUs https://review.openstack.org/385024 | |
| 15:54:27 | openstackgerrit | sahid proposed openstack/nova: libvirt: make emulator threads to run on the reserved pCPU https://review.openstack.org/400878 | |
| 15:54:28 | openstackgerrit | sahid proposed openstack/nova: numa: take into account cpus reserved https://review.openstack.org/400876 | |