Tuesday, 2022-06-28

opendevreviewOpenStack Proposal Bot proposed openstack/glance master: Imported Translations from Zanata  https://review.opendev.org/c/openstack/glance/+/84684502:37
*** abhishekk is now known as akekane|home04:48
*** akekane|home is now known as abhishekk04:48
opendevreviewAbhishek Kekane proposed openstack/glance master: [APIImpact] Correct API response code for DELETE cache APIs  https://review.opendev.org/c/openstack/glance/+/84787705:46
abhishekkdansmith, croelandt, jokke_  ^^05:49
abhishekkcroelandt, dansmith jokke_ if you are around kind look at new glance patch, https://review.opendev.org/c/openstack/glance/+/84787713:36
dansmiththat was back in xena?13:38
abhishekkyoga13:39
dansmithoh okay the spec you linked said xena13:41
abhishekkspec was in xena but it is implemented in yoga13:41
dansmithack13:44
dansmithso just thinking about this a sec.. 200->204 seems kinda "meh" in terms of usefulness,13:45
dansmithbut also, this is a similar case to the PUT right? you ask for it to be deleted, but we don't do it until later right?13:45
abhishekkyes13:45
dansmithjust saying, 200->202 for the PUT is pretty important to say "no, we didn't already do that, we're going to do it sometime later", but delete is less important anyway, but also 200->204 doesn't tell us anything else really useful13:47
dansmithso I assume the same justification for the PUT change still applies, but the reasoning is weaker for this I think.13:47
abhishekkSo can we amend the spec and api reference doc and keep it 200 only?13:47
dansmithmaybe we see what others think? too bad rosmaita is on vacation, I'd be interested in his opinion13:48
dansmithyeah, could do that, let's see what jokke_ and croelandt think13:48
abhishekkhmm, because our api reference is also saying 204 is valid response code for these APIs13:48
dansmithmeaning the apiref, code and spec all disagree?13:49
abhishekkspec and api reference is 204 and code is 20013:49
dansmither, right13:49
dansmithwell, what's in the code is what will break people, but yeah13:50
dansmithgood that you noticed the disparity regardless13:50
dansmithlooks like the other DELETEs in the glance api also specify 20413:51
abhishekkyes13:51
abhishekkit wasn't me, it was pointed out by martin kopec on my tempest patch13:52
dansmithah, heh13:53
abhishekkBrian will not be here until next Wednesday, i guess jokke_ is also not around 13:54
dansmithyeah I know rosmaita is out13:57
dansmithI think making sure jokke_ is okay with this is a good idea,13:57
dansmithbecause like last time, it's an API change, but the justification is much weaker for this, IMHO13:57
dansmithso I commented with some of the discussion above, I assume he'll be around at some point13:58
dansmithis that cool?13:58
abhishekkack, only thing is we need to be consistent 13:58
abhishekkno worries13:58
dansmithwhat's the answer to the version question? do we consider the existing bump for PUT to cover this?13:59
dansmithand I assume you'll backport this as well?13:59
abhishekkyes13:59
abhishekkIMO we can consider existing version bump to cover this14:00
dansmithokay14:00
abhishekkbecause we haven't released anything yet14:00
dansmithack14:01
croelandtyeah I'd kind of like to see a version bump and a release note15:03
croelandteven though I doubt this will break anything15:04
dansmithwell, abhishekk says we're covered under the other bump we just made for the PUT15:05
dansmithbut a reno, yeah15:05
abhishekkI will add a reno once we decided to go ahead with this15:06
croelandtabhishekk: can you +W https://review.opendev.org/c/openstack/glance-specs/+/734683 ?15:06
abhishekklooking15:06
abhishekkyep15:06
opendevreviewMerged openstack/glance-specs master: Update proposal for duplication image download  https://review.opendev.org/c/openstack/glance-specs/+/73468315:13
jokke_abhishekk: dansmith: I'm almost leaning towards altering the doc instead. 200 with "Deleted." would be equally right response [0]. I really do like consistency 'though which is IMO the only excuse why we should do that change https://www.rfc-editor.org/rfc/rfc9110.html#name-delete17:19
jokke_Or rather "Invalidated" 17:19
dansmithyes, I agree, 204 is no more correct than 200 for this, although I think 202 *is* since we queue the delete (right?)17:20
dansmithconsistency is important, and this delete is pretty limited use and scope (like we said for PUT) which also makes fixing it not a big deal I think17:20
dansmithso I think I'd lean towards fixing it since ALL the other deletes are 204, but not super opinionated17:21
jokke_The cache deletes are actually synchronous so 202 would be wrong17:21
dansmithah okay17:21
dansmithI guess the other argument for fixing it,17:22
jokke_that's why I'm bit torn between consistency vs API stability for consistency :P17:22
dansmithis we already fixed the PUT, so one version that covers the messed-up return values for two related calls is reasonable symmetry17:22
dansmithit's not like we'd need two versions, or that the two corrections are wildly unrelated17:22
dansmiththey're two calls for the same API, same feature, same kinda original oversight17:23
jokke_there is that17:23
jokke_well like abhishekk said ... the previous one is not released so if someone is relying on that already, they would need to change their caching code anyways17:24
dansmithyeah17:24
dansmithI suspect brian would be pro (maybe strongly) for just fixing it.. abhishekk clearly is, and I'm mildly for fixing it... are you opposed?17:25
* jokke_ wishes HTTP had just one 200 "got ya"17:25
abhishekkright17:25
dansmithwell, in reality, most things return 200 regardless, so there kinda is :P17:25
abhishekkI am pro for fixing it17:26
jokke_mabe we should too ... little sed and everything 2XX -> 20017:26
jokke_:P17:26
jokke_but no I'm not really opposed. Kind of feeling that we shat the bed already so we might as well :D17:27
dansmiththere seems to be more weight on the side of fixing it while we have this chance, so I say we go for it17:27
dansmithhah, right, my feeling exactly :)17:27
abhishekk++17:27
jokke_abhishekk: could you please just make sure that these are all then ;)17:27
abhishekklet me add quick release notes17:27
dansmithabhishekk: also please reno17:27
dansmithyeah17:27
abhishekkyep PUT is already fixed17:27
abhishekkand list one is correct17:27
abhishekklist --> GET is 20017:28
jokke_good, we don't return 202 with the list there "Listing these, but there might be more, you'll never know"17:28
dansmithreturn 200 + randint(0, 99)17:29
jokke_++17:29
jokke_on everything17:29
abhishekk:D17:33
opendevreviewAbhishek Kekane proposed openstack/glance master: [APIImpact] Correct API response code for DELETE cache APIs  https://review.opendev.org/c/openstack/glance/+/84787717:35
abhishekkdansmith, jokke_ ^17:35
jokke_We should also add default body to all our 200s that doesn't have one specified 200 "Or something there abouts, it'll be grand anyways"17:37
jokke_even better 200 "Ish, you're grand"17:38
abhishekk:D17:39
abhishekkjokke_, dansmith if patch gets merged today/tonight can either of you please propose backport to stable/yoga17:41
dansmithabhishekk: we can propose it now, right?17:42
dansmithassuming no rebase needs to happen, the hash will be the same17:42
abhishekkwe can as there is no patch in queue to be merged17:42
opendevreviewAbhishek Kekane proposed openstack/glance stable/yoga: [APIImpact] Correct API response code for DELETE cache APIs  https://review.opendev.org/c/openstack/glance/+/84800117:43
abhishekk^^17:43
* dansmith was git reviewing already17:43
dansmithI interrupted it ;)17:43
abhishekkahh17:43
dansmithabhishekk: note that I can't +2 that, so I guess my work here is done :P17:44
abhishekksorry :D17:44
dansmithno complaints from me :)17:44
abhishekkI will abandon mine, you can propose backport17:44
dansmithnonono17:44
abhishekkok17:44
dansmithoh, if you want me to so you can vote?17:44
abhishekkyes17:44
opendevreviewDan Smith proposed openstack/glance stable/yoga: [APIImpact] Correct API response code for DELETE cache APIs  https://review.opendev.org/c/openstack/glance/+/84800117:44
abhishekkcool, abandoned mine17:45
* abhishekk signing out for the day18:00

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!