Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-09-20
14:29:49 openstackgerrit LIU Yulong proposed openstack/openstacksdk master: Support firewall service for SDK https://review.openstack.org/592303
15:08:19 mnaser mordred: i remember you mentioning there was a way to get an 'http client' for the placement service using openstacksdk :(
15:08:23 mnaser but i cant remember where or how
15:09:03 mnaser it was something that even if there is no placement support, doing something got you an http client
15:18:11 mordred mnaser: conn.placement should be an http client for placement
15:18:44 mordred mnaser: so conn.placement.get('/some/url') should return you a Response object
15:18:55 mnaser mordred: wonderful, i'll give that a shot. writing tooling to do an audit of placement/nova
15:19:33 mordred cool! we should probably add real placement support
15:19:46 mordred maybe one day I'll sucker cdent into doing that
15:19:51 mnaser mordred: mriedem made osc-placement so it cant be that much work
15:20:14 cdent andrey and a few others did a lot of that work
15:20:24 cdent mordred: curl and/or gabbi 4evah
15:23:42 mnaser mordred: http://paste.openstack.org/show/730447/ shucks
15:23:45 mnaser i remember us talking about this
15:23:53 mnaser maybe the fix wasn't released
15:24:04 mnaser cdent pushed up something about this too
15:25:51 mriedem yeah i only reviewed osc-placement...
15:26:11 cdent mnaser: you thinking of this: https://review.openstack.org/#/c/575117/ ?
15:26:32 mriedem cdent: no,
15:26:34 mnaser cdent: ah yes
15:26:34 mriedem more about,
15:26:47 mriedem oh - i think what mnaser is looking for is a tool to look for orphans in placement
15:26:47 mnaser this is a queens deployment
15:27:02 mriedem e.g. orphaned providers or allocations for things that don't actually exist in nova anymore
15:27:05 mriedem like compute nodes and instances
15:27:09 mnaser well that's what im trying to write if i get a working client :p
15:27:29 cdent it's fun to "yes" and "no" to the same thing
15:27:31 mriedem personally i think that should live in nova-manage placement *
15:27:51 mordred mnaser: ah - that hasn't landed yet :(
15:27:53 mnaser i dont mind adding that in
15:28:02 mriedem but as i said yesterday, we can't know if the providers/allocations in placement are for nova things really without consumer types
15:28:16 mriedem although we can probably glean that from knowing if the provider has VCPU inventory
15:28:35 mordred oh - wait - I can't read - it did land
15:28:39 mriedem (1) is it root (2) does it have vcpu inventory, then it's likely a compute node
15:28:54 mnaser mordred: landed on the API side in rocky, i dont think sdk works around this somehow locally
15:29:35 mordred mnaser: yah - I do not believe it has any local workarounds. we could maybe make one? I think the best workaround would be to put in a placement_endpoint_override
15:29:39 mnaser mriedem: i was going to assume it was a compute node all the time (because in my use case it always is) but i guess adding those stuff can be trivial
15:29:56 mordred in your clouds.yaml or whatever - so that we don't try to run discovery and fail due to lack of status
15:30:16 mnaser im totally using env variables out of lazyness
15:30:22 mnaser so i assume OS_PLACEMENT_ENDPOINT_OVERRIDE ?
15:30:46 mordred yah
15:32:24 cdent mriedem: speaking of "does it have vcpu inventory" I made https://review.openstack.org/#/c/600016/ in response to digging in some of the placement-db using nova-manage and nova-status stuff
15:33:23 mnaser mordred: yay that worked, i'll workaround with that and then maybe i'll ask for help from mriedem to bring that into nova-manage
15:33:36 mordred coolio
15:33:40 mnaser merci beaucoup
15:33:44 mordred \o/
17:04:28 edleafe cdent: elmiko: dtantsur: https://etherpad.openstack.org/p/api-sig-newsletter is ready for your review
17:05:12 dtantsur looks good
17:07:47 cdent "has finally realized" more like "accepted" but yeah: looks good
17:08:51 elmiko edleafe: ++ lgtm
17:08:58 edleafe cdent: we all realized that about you much earlier :)
17:48:03 elmiko edleafe cdent , i enjoyed your writeups from the ptg, thanks!
18:02:18 edleafe dtantsur|afk: wish you could have been there too
#openstack-sdks - 2018-09-21
00:57:14 openstackgerrit LIU Yulong proposed openstack/openstacksdk master: Support firewall service for SDK https://review.openstack.org/592303
13:45:55 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Use discovery instead of config to create proxies https://review.openstack.org/600292
13:57:15 Shrews +306, -1600
13:57:18 Shrews wow
13:57:29 Shrews i can't wait to fully review that
13:57:40 Shrews while drinking
14:00:19 mordred Shrews: one of these days it's going to pass the bifrost job and I'm going to be happy
14:01:49 mordred Shrews: for now I've put in an autohold so I can poke at it again
14:02:04 mordred Shrews: I *thought* I'd figured out the issue 2 days ago ... sigh
14:11:01 Shrews mordred: i'm sort of waiting for all green before i look at it
14:11:15 mordred it's the best bet
14:12:33 dtantsur on one hand, bifrost is pita :) on the other - it's the only thing that apparently provides this sort of coverage
14:14:54 mordred dtantsur: yah - I've been adding unit tests as I figure out issues - but apparently there's still one I don't have :
14:14:56 mordred :(
14:15:37 mordred dtantsur: it's been really useful to sort out the no-catalog versioned and unversioned endpoint use-case :)
14:17:22 dtantsur oh yeah
14:17:49 dtantsur mordred: unrelated: do you plan on openstacksdk release any soon? I could give switching metalsmith to it a try
14:20:37 mordred dtantsur: yes - I think we're ready for one - but I was gonna see if I could get that discovery patch landed
14:20:49 dtantsur right
14:21:13 mordred dtantsur: but if it doesn't turn green today, maybe we'll cut a release with what we have so far on monday
14:34:50 dtantsur yeah, and then another one. release numbers are cheap :)
14:43:32 mordred ++
14:48:04 mordred Shrews: if you're bored and feel like flaying your brain a little ... https://review.openstack.org/#/c/603739/
14:48:41 Shrews k k
14:50:35 mordred Shrews: I think we might want corvus to review it when he gets back as well - since it's a fundamental change to the internals of the task manager - but I *think* it's solid
14:51:19 Shrews passed pep8... ship it
14:52:49 mordred ++
15:49:30 Shrews mordred: why the call to ret.exception()? We don't seem to do anything with it
15:50:15 Shrews ret.result() should raise an exception if one occurred in the call (if i read the Futures docs correction)
15:50:23 Shrews s/correction/correctly/
15:54:10 mordred Shrews: oh - I couldn't find mention that result would do that - so I guess it's senseless
15:54:36 Shrews mordred: https://docs.python.org/3/library/concurrent.futures.html#concurrent.futures.Future.result
15:55:28 mordred Shrews: so, whatyou're saying is thatif I could actually read simple english, my life would be better?
16:07:23 Shrews mordred: left comments :-P
16:08:43 mordred Shrews: \o/
16:09:31 Shrews mordred: i'm not sure if anyone actually uses submit_function_async(), but probably a bad idea to totally remove an API call
16:09:38 Shrews correct me if i'm wrong
16:09:49 Shrews 'cause i know nothing about anything
16:09:54 Shrews 'cause Friday
16:10:17 mordred Shrews: yeah - I agree with you in principle, but in practice I'm 99.9% sure that nodepool is the only consumer of anything related to taskmanager
16:10:29 Shrews i checked nodepool for it, fwiw
16:10:31 Shrews not used
16:10:34 mordred yah
16:10:39 Shrews so i can be convinced to remove it
16:11:01 Shrews mmmm, frothy
16:18:10 openstackgerrit Matthias Lisin proposed openstack/openstacksdk master: Support firewall service for SDK https://review.openstack.org/592303
16:28:27 mordred dtantsur: I think I found the issue with the bifrost

Earlier   Later