| *** yikun has joined #openstack-placement | 01:44 | |
| *** belmoreira has joined #openstack-placement | 06:51 | |
| openstackgerrit | yatin proposed openstack/placement master: Make placement base API return version without auth https://review.opendev.org/674543 | 07:39 |
|---|---|---|
| *** belmoreira has quit IRC | 08:08 | |
| *** belmoreira has joined #openstack-placement | 08:12 | |
| *** tssurya has joined #openstack-placement | 08:41 | |
| *** e0ne has joined #openstack-placement | 08:48 | |
| *** belmoreira has quit IRC | 09:03 | |
| *** ykarel has joined #openstack-placement | 09:11 | |
| ykarel | cores can u look at https://review.opendev.org/#/c/674543/ | 09:13 |
| *** takashin has left #openstack-placement | 09:32 | |
| *** belmoreira has joined #openstack-placement | 09:49 | |
| *** cdent has joined #openstack-placement | 09:49 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make placement base API return version without auth https://review.opendev.org/674543 | 10:33 |
| ykarel | cdent, Thanks for ^^ | 11:08 |
| cdent | ykarel: you're welcome | 11:08 |
| cdent | I suspect that'll get merged pretty soon. Is it something you're hoping to be able to backport? | 11:08 |
| ykarel | cdent, not looking for backport, only seen this issue after nova patch merged which swictched to openstacksdk | 11:09 |
| ykarel | but it's seems a good candidate for backport | 11:09 |
| cdent | ah, okay, cool | 11:09 |
| ykarel | cdent, but i still can't get what part is appending / to requests in devstack | 11:10 |
| cdent | yeah, I'll ask around with a few other folk | 11:10 |
| ykarel | okk Thanks | 11:10 |
| cdent | it might be mod_proxy_uwsgi | 11:10 |
| ykarel | yup i am guessing that | 11:10 |
| ykarel | but can't find a reference | 11:10 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Track usage info on RequestWideSearchContext https://review.opendev.org/674581 | 12:00 |
| cdent | efried: that is riffing off you riffing off me | 12:01 |
| *** gibi_off is now known as gibi | 12:01 | |
| cdent | gibi: you back? welcome back! | 12:02 |
| gibi | cdent: thanks | 12:02 |
| gibi | now I have a looong list of mails to process :D | 12:02 |
| cdent | i bet | 12:03 |
| *** tssurya has quit IRC | 12:07 | |
| *** tssurya has joined #openstack-placement | 12:08 | |
| *** jroll has quit IRC | 12:39 | |
| *** belmoreira has quit IRC | 12:39 | |
| *** jroll has joined #openstack-placement | 12:39 | |
| *** belmoreira has joined #openstack-placement | 12:59 | |
| efried | o/ gibi! | 13:10 |
| efried | we missed you :* | 13:10 |
| gibi | efried: o/ | 13:10 |
| gibi | :) | 13:10 |
| *** mriedem has joined #openstack-placement | 13:14 | |
| *** ykarel is now known as ykarel|afk | 13:22 | |
| efried | cdent: Will https://review.opendev.org/#/c/674543/ actually help with https://bugs.launchpad.net/nova/+bug/1838807 ? | 13:28 |
| openstack | Launchpad bug 1838807 in OpenStack Compute (nova) "UnsupportedServiceVersion: Service placement has no discoverable version. The resulting Proxy object will only have direct passthrough REST capabilities." [Medium,In progress] - Assigned to Takashi NATSUME (natsume-takashi) | 13:28 |
| cdent | efried: yes, see my comment | 13:28 |
| cdent | I think the issue is fairly...weird | 13:29 |
| efried | I asked mordred to put in a (passthrough) Proxy in sdk for Placement | 13:29 |
| efried | ...which I thought was what was missing, and mordred seemed to agree that that would sort the problem. | 13:29 |
| cdent | actually | 13:29 |
| cdent | hmmm | 13:29 |
| cdent | I'm not sure | 13:29 |
| cdent | there is always a discoverable version | 13:29 |
| cdent | but if you request wrong, you won't get it | 13:30 |
| efried | I agree having version endpoint support fixed as you've done is a necessary thing. | 13:30 |
| cdent | see the other bug for a bit more context | 13:30 |
| cdent | it's possible that somet hings are request /placement and getting something weird, which is not triggering an error in sdk (but probably should be) but telling it couldn't find version doc | 13:31 |
| cdent | but there is (and always has been) a version doc | 13:31 |
| openstackgerrit | Matt Riedemann proposed openstack/placement master: Make placement base API return version without auth https://review.opendev.org/674543 | 13:33 |
| efried | Yeah, I did this last week: | 13:35 |
| efried | 3:15:18 PM - efried: […] curl http://192.168.218.27/placement | 13:35 |
| efried | 3:15:18 PM - efried: {"versions": [{"status": "CURRENT", "min_version": "1.0", "max_version": "1.36", "id": "v1.0", "links": [{"href": "", "rel": "self"}]}]} | 13:35 |
| efried | so like, at least curl is happy with non-trailing-slash | 13:36 |
| efried | another piece of the discussion was: should placement be using versioned endpoints? .../placement/v1/* | 13:36 |
| cdent | efried: it makes a big big difference whether you are running in devstack, tripleo, something else | 13:37 |
| efried | okay. Above was devstack | 13:37 |
| cdent | placement is and always will be unversioned, I've had that discussion with mordred before | 13:37 |
| cdent | if it needs to have a new version, then it needs to be a new service | 13:38 |
| cdent | and that's fine | 13:38 |
| efried | okay. | 13:38 |
| cdent | they are just urls | 13:38 |
| cdent | but versions in urls is a sing | 13:38 |
| cdent | sin! | 13:38 |
| cdent | I believe ykarel|afk (from earlier, the original author on that fix) has more details on which deployments a / or '' happens | 13:38 |
| cdent | also, efried, I'm ignoring your mock bait, but almost got distracted going on a screed, decided my time was spent elsewhere | 13:39 |
| efried | :P | 13:40 |
| efried | I was of course totally baiting you. Not even slightly serious. | 13:40 |
| ykarel|afk | efried, so if u run curl http://192.168.218.27/placement | 13:42 |
| ykarel|afk | also check logs | 13:42 |
| ykarel|afk | there you will find actual request was GET /placement/ | 13:42 |
| ykarel|afk | checked in a devstack env ^^ | 13:42 |
| ykarel|afk | so something is adding / with devstack setup, so 1st thing placement should fix it, second openstacksdk should provide compatibility with urls not having / | 13:43 |
| ykarel|afk | as earlier with keystone adapter it was working, i browsed through keystoneauth code, and there the are hacking urls multiple places | 13:44 |
| efried | ykarel|afk: the latter should be brought up with mordred in the #openstack-sdks channel | 13:44 |
| efried | Yes, this is one of the reasons we wanted to merge this cutover to sdk sooner rather than later. To flush out weirdnesses like this. | 13:44 |
| efried | sdk is piggybacking on ksa by the way. | 13:44 |
| ykarel|afk | efried, sure i can ping him later, currently need to go out | 13:45 |
| efried | but I'm sure some pieces are being sidestepped. | 13:45 |
| *** ykarel|afk is now known as ykarel|away | 13:52 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Add RequestWideSearchContext.summaries_by_id https://review.opendev.org/674254 | 14:01 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Further optimize _build_provider_summaries https://review.opendev.org/674349 | 14:01 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Track usage info on RequestWideSearchContext https://review.opendev.org/674581 | 14:01 |
| cdent | efried: all that ^ needed a rebase to catch up with change in lower-constraits | 14:01 |
| efried | oh, I thought I had done that. | 14:01 |
| *** belmoreira has quit IRC | 14:02 | |
| cdent | I might have messed it up when I added another patch to the end | 14:02 |
| efried | k | 14:02 |
| efried | I'm getting set up to look at the last patch soonisg. | 14:02 |
| efried | soonish | 14:02 |
| *** belmoreira has joined #openstack-placement | 14:03 | |
| efried | (/me got in the poison ivy again, on antihistamines, brain furry) | 14:03 |
| cdent | it's interesting and probably opens up a fair few other possibilities | 14:03 |
| cdent | oops | 14:03 |
| cdent | something got me yesterday | 14:03 |
| cdent | i've kind of reached a stopping point on the profiling. I can't really make any further headway until there's a new common base for analysis | 14:09 |
| cdent | gibi: if you've got a few cycles to spare reviewing the several performance-related patches I made to placement near the end of last week, that would be awesome | 14:10 |
| gibi | cdent: I put it on my list | 14:23 |
| cdent | gibi: thanks! | 14:23 |
| gibi | cdent: I've already seen your work using _slots_ | 14:23 |
| cdent | are not because of missing rebase. "'expanding' parameters with an empty list not supported until SQLAlchemy 1.3" | 14:24 |
| cdent | i'll fix that | 14:24 |
| gibi | mriedem, cdent: btw. regarding profiling the nova scheduler with os-profile vs. cProfiler. I think if we targeting a single process then cProfiler is better. os-profiler is more for profiling a multi process, multi module interaction as it has pretty severe overhead to store each trace point | 14:25 |
| gibi | if I remember correctly neutron guys see the rally tests slowing down 5 times when os-profiler is enabled | 14:26 |
| cdent | yow | 14:26 |
| mriedem | gibi: ack thanks. i at least got a patch to add a nova osprofiler job to the experimental queue https://review.opendev.org/#/c/674326/ and from looking at some of that found a thing to clean up https://review.opendev.org/#/c/674422/ but it's hitting some weirdness in a libvirt numa functional test | 14:27 |
| mriedem | totally unrelated to the -placement channel though :) | 14:27 |
| mriedem | also, welcome back | 14:27 |
| gibi | mriedem: thanks. I've added your patches to my queue too :) | 14:28 |
| * gibi building backlog today | 14:29 | |
| *** mriedem has quit IRC | 14:29 | |
| *** mriedem has joined #openstack-placement | 14:29 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Further optimize _build_provider_summaries https://review.opendev.org/674349 | 14:33 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Track usage info on RequestWideSearchContext https://review.opendev.org/674581 | 14:33 |
| cdent | efried: on "further optimize" had to add a skip if no rp_ids | 14:33 |
| efried | cdent: You gonna still +2 that guy? | 14:39 |
| cdent | efried: I was waiting to ask you if you thought my little fix took me out of the running | 14:39 |
| cdent | 2nd opinion all that | 14:40 |
| efried | cdent: Add my +1 and go ahead and +2 the thing | 14:40 |
| cdent | aye | 14:40 |
| cdent | RPCandidates.filter_by_rp is now (interestingly) a noticeable contributor | 14:43 |
| *** belmoreira has quit IRC | 14:48 | |
| *** belmoreira has joined #openstack-placement | 14:51 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make _get_trees_with_traits return a set https://review.opendev.org/674630 | 15:03 |
| cdent | efried: becasue you're not already busy enough ^ | 15:03 |
| cdent | we appear to be down under 2 seconds from somewhere above 6 | 15:11 |
| cdent | but that's only guesswork | 15:11 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Make _get_trees_with_traits return a set https://review.opendev.org/674630 | 15:11 |
| efried | cdent: that's with the complex-y topo? | 15:12 |
| cdent | yeah, I have my database full of about 6500 of the complexy providers (plus some others) | 15:12 |
| cdent | 9423 total providers in the thing | 15:13 |
| cdent | 6551 in the results of a GET /a_c with the complexy topo | 15:14 |
| cdent | that's just with that stack | 15:14 |
| cdent | still go the double join to get rid of | 15:15 |
| cdent | got | 15:15 |
| cdent | and another one of the redudant casts | 15:15 |
| cdent | once we get all that merged it might make sense to do some small scale profiling, because now things like transforming all that data to json is showing up as a big factor | 15:16 |
| cdent | and stuff in the creation of the data is getting obscured | 15:16 |
| *** belmoreira has quit IRC | 15:52 | |
| *** belmoreira has joined #openstack-placement | 16:02 | |
| *** belmoreira has quit IRC | 16:03 | |
| *** e0ne has quit IRC | 16:19 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Use expanding bindparam in extend_usages_by_provider_tree https://review.opendev.org/674647 | 16:21 |
| openstackgerrit | Merged openstack/placement master: Make placement base API return version without auth https://review.opendev.org/674543 | 17:07 |
| *** tssurya has quit IRC | 17:34 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Use orjson in python3 for allocation candidate dump https://review.opendev.org/674661 | 17:49 |
| *** ykarel|away has quit IRC | 17:56 | |
| *** e0ne has joined #openstack-placement | 18:05 | |
| *** e0ne has quit IRC | 18:45 | |
| *** cdent has quit IRC | 18:59 | |
| *** e0ne has joined #openstack-placement | 19:16 | |
| *** e0ne has quit IRC | 19:31 | |
| *** mriedem has quit IRC | 21:19 | |
| *** mriedem has joined #openstack-placement | 21:21 | |
| *** artom has joined #openstack-placement | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!