Tuesday, 2016-01-12

*** dims_ has joined #openstack-release00:39
*** dims has quit IRC00:40
*** dims_ has quit IRC01:48
*** dims has joined #openstack-release01:48
*** dims has quit IRC01:54
*** jhesketh has quit IRC02:05
*** jhesketh has joined #openstack-release02:05
*** dims has joined #openstack-release02:06
*** amotoki has joined #openstack-release02:54
*** amotoki has quit IRC03:12
*** amotoki has joined #openstack-release03:14
*** sigmavirus24_awa is now known as sigmavirus2403:22
*** dims has quit IRC03:24
*** doug-fish has joined #openstack-release03:30
*** amotoki has quit IRC03:36
*** doug-fish has quit IRC03:58
*** amotoki has joined #openstack-release04:04
*** sigmavirus24 is now known as sigmavirus24_awa05:18
*** dtantsur|afk is now known as dtantsur08:57
*** amotoki has quit IRC09:05
*** amotoki has joined #openstack-release09:07
*** amotoki has quit IRC09:07
*** openstackgerrit has quit IRC09:17
*** openstackgerrit has joined #openstack-release09:17
*** sileht has quit IRC09:52
*** amotoki has joined #openstack-release09:59
*** sileht has joined #openstack-release10:03
*** dims has joined #openstack-release11:05
*** openstackgerrit has quit IRC11:17
*** openstackgerrit has joined #openstack-release11:17
*** amotoki has quit IRC11:46
*** amotoki has joined #openstack-release11:50
*** amotoki has quit IRC12:02
*** mriedem_away has quit IRC13:15
*** sigmavirus24_awa is now known as sigmavirus2413:31
*** sigmavirus24 is now known as sigmavirus24_awa13:40
*** BobBall has joined #openstack-release13:51
BobBalldhellmann / ttx : Any chance you could look at https://review.openstack.org/#/c/264741/ ?  ryu 3.29 is incompatible with oslo.config 3.1+ and it's causing Citrix's internal CI to fail.  We'd like to prevent ryu 3.29 from being used (as commented on the review, ryu have a fix in place pending the next release, so excluding 3.29 should be sufficient)13:54
ttxBobBall: looking13:54
ttxlooks goof13:57
ttxor good13:57
BobBallAwesome - thanks13:58
*** gordc has joined #openstack-release14:04
*** amotoki has joined #openstack-release14:20
*** amotoki has quit IRC14:21
*** mriedem has joined #openstack-release14:25
*** amotoki has joined #openstack-release14:40
*** amotoki has quit IRC14:47
*** doug-fish has joined #openstack-release15:02
openstackgerritSteve Martinelli proposed openstack/releases: release keystoneauth and keystoneclient for stable/liberty  https://review.openstack.org/26643315:24
stevemardhellmann: dims bknudson_ ^15:24
dimsstevemar : will wait for the jobs to run (need to look at output from list-changes)15:26
stevemardims: cool cool15:27
stevemardims: for stable releases, we don't need to make a change to requirements right?15:27
dimsstevemar : y, we leave those alone for now15:30
openstackgerritgordon chung proposed openstack/releases: ceilometerclient 2.2.0  https://review.openstack.org/26643815:32
dhellmannBobBall, ttx: approved15:46
dhellmannstevebaker : looking15:47
dhellmannstevebaker : oops, stevemar15:47
dhellmannstevemar: you're missing send-announcements-to in those files http://logs.openstack.org/33/266433/1/check/gate-releases-tox-validate/7c1f454/console.html#_2016-01-12_15_43_28_17315:47
BobBallAwesome - thanks dhellmann!15:48
ttxdhellmann: interesting discussion about release models at the stable team meeting, you should review the logs when done15:50
ttxwe may want to better document the consequences of picking "independent" wrt testing15:50
dhellmannttx: noted15:53
dhellmannI should add that meeting to my calendar15:53
ttxdhellmann: link: http://eavesdrop.openstack.org/meetings/stable/2016/stable.2016-01-12-15.01.log.html#l-3815:54
openstackgerritSteve Martinelli proposed openstack/releases: release keystoneauth and keystoneclient for stable/liberty  https://review.openstack.org/26643315:56
openstackgerritDoug Hellmann proposed openstack/releases: copy send-announcements-to from mitaka to kilo and liberty  https://review.openstack.org/26644815:59
dhellmannttx: we should get ^^ in to avoid causing issues for folks with stable release requests15:59
ttxdhellmann: looking16:06
dhellmannttx: it sounds like mriedem and sdague were proposing that release:independent projects can't be co-gated with release:cycle-with-intermediary or milestone-based projects16:09
dhellmannthe only real reason I see for that is the implementation detail of our d-g jobs, though, right?16:10
ttxI think it goes beyond that, but I won't talk for them ;)16:11
mriedemdhellmann: well i think that's how ceilometer is getting around it right now16:11
dhellmannI mean there's the conceptual issue of saying "what is stable" but the roadblock that makes having lots of cross testing hard is the implementation of figuring out what branch to check out when16:12
ttxit's also about how likely you are to break backwards compat too16:12
dhellmannis it? why does having more stable branches imply more breaks?16:13
ttxmore stable branches implies less breaks :)16:13
dhellmannok, that's what I would have thought, too, I guess I misinterpreted what you said16:14
dhellmannthe fact that we have to do so much fooling around to configure jobs specially when we have branches with mis-matched names is an implementation detail we can fix, imo16:15
sdaguedhellmann: it's also about being able to communicate to users16:15
sdaguewhat are they supposed to run with their liberty cloud?16:15
dhellmannthe version(s) we test with. I'm not sure why a list of 2-3 is that much harder than 116:15
dhellmannI mean, I understand the value of having a single stable branch per project, too, I'm just trying to understand why we want to add a new rule in a case where we have more than one16:17
sdaguemaybe, I feel like it adds a pretty big level of complexity to an already complex system16:17
dhellmannis there something wrong with the ceilometer/gnocchi situation or is it just a pain for the developers who have chosen to do something outside of our norms?16:17
sdagueit has impacted beyond just the projects that deviated, because of co-gating16:18
dhellmannwith lifeless' stability changes to libs, we're going to end up having tests for lower bounds of requirements as well as upper16:18
sdagueand has managed to block other things16:18
*** sigmavirus24_awa is now known as sigmavirus2416:18
sdaguethis is only about co-gating branches16:18
dhellmannok, so why is gnocchi being installed from source in projects other than ceilometer?16:18
sdaguebecause ceilometer co-gates with other things16:19
dhellmannI'm OK with us saying "don't do this", too, fwiw, but I'd like to understand the real issue16:19
dhellmannbut if gnocchi is a dependency of ceilometer, shouldn't it be installed from packages in those other tests?16:20
sdagueso when this all goes sideways, tempest can't land any patches, for instance16:20
sdagueexcept they set it up to co-gate iirc16:20
dhellmannwho set what up? the telemetry team set up cogating between gnocchi and projects other than ceilometer?16:21
dhellmannhow does that affect tempest's ability to land patches? I only see other telemetry team projects there16:21
sdagueok, well, I only have about half the problem in my head because I've mostly been seeing some of the breaks from a far16:22
sdagueI'm happy to walk away from this one16:22
dhellmannit's simple to create a rule that release:independent projects cannot co-gate with other projects. it's easy to understand, and maybe gets us what we want in terms of managing complexity and avoiding one project breaking the gate in unexpected ways.16:24
dhellmannotoh, this team seems to have a reason for wanting to do this, so I'd like to understand the goals and real ramifications before supporting a change like that.16:25
*** BobBall has quit IRC16:25
dhellmanngordc : ^^16:26
mriedemthe ceilometer stable/liberty thing doesn't break tempest, it breaks the dsvm integration job for ceilometer16:38
mriedemceilometer isn't in the integrated gate jobs16:38
mriedemdhellmann: and i'm told gnocchi isn't a library, it's a service, so that's why they install from master16:39
mriedemceilometer only depends on python-gnocchiclient16:39
mriedemheh, or not https://github.com/openstack/ceilometer/blob/stable/liberty/requirements.txt16:40
* gordc reading16:43
*** dtantsur is now known as dtantsur|brb16:45
mriedemgordc: if ceilometer doesn't require gnocchiclient in stable/liberty, why is it needed in the stable/liberty job (gnocchi that is)16:47
mriedemgordc: oh i guess it is https://github.com/openstack/ceilometer/blob/stable/liberty/ceilometer/tests/integration/hooks/post_test_hook.sh#L4316:49
gordcmriedem: we don't add drivers to our requirements16:49
gordcbut yes, that gate test specifically tests the integration between ceilometer+gnocchi+aodh+heat16:50
gordc+nova i guess16:50
mriedemgordc: ok, the master gnocchi + stable/liberty ceilometer thing came up in the stable team meeting this mornign16:50
gordcso just background on why gnocchi doesn't have same branches as all the other projects... it started off as it's own and has uses cases beyond openstack.16:50
mriedemgordc: so i see gnocchi is in pypi,16:51
mriedemwhy doesn't ceilometer on stable/liberty just install from pypi?>16:51
gordcmriedem: yep16:51
gordcmriedem: we actually could (i believe). we were just trying to follow all the gating protocols/"standards"16:52
mriedemexcept gnocchi doesn't follow those16:52
gordcjd__: ^16:52
gordcmriedem: because we name our stables differently?16:53
mriedemyeah, well and https://github.com/openstack-infra/project-config/blob/d2cfff4803dba96f2a5abaf82f4e3b5d68f767a4/jenkins/jobs/ceilometer.yaml#L21416:53
gordcwell we were trying to use enable_plugin which should allow you to choose branches16:53
mriedemthat means you can use 3 stable gnocchi branches for ceilometer stable/kilo testing rihgt?16:53
gordcyes. those are valid for kilo16:54
mriedemi guess it's confusing from a downstream user pov16:54
mriedemwhich version of gnocchi should i run with ceilometer stable/liberty16:54
mriedemif ceilometer doesn't put a version constraint on gnocchi16:54
gordci would think it's similar to any other driver. ie what version of mysql should you run?16:55
mriedemif gnocchi is more than just openstack, what's it's backward compat policy? just semver?16:55
mriedemthinking out loud, if gnocchi 2.0 drops some things that ceilometer stable/liberty were using, the dsvm integration job shouldn't break b/c it should be using gnoccho stable/1.3 branch16:57
gordci understand it complicates stuff because of naming differences, but our idea is gnocchi should mimic how we treat other db. it shouldn't be "stable/liberty" only because then it only makes senes in oepnstack context.16:57
jd__we could pip install it, I indeed already proposed that16:57
jd__we just didn't implement it16:58
mriedemi feel like we should pip install it16:58
gordcthe client or both?16:58
ttxthat would make the version pairing more explicit16:59
ttxbecause I don't expect that many people to read jenkins/jobs/ceilometer.yaml#L21417:00
gordcmriedem: sorta/not-related but the project-config change worked for liberty17:00
mriedemgordc: ceilometer stable/liberty isn't using the client - there is a todo in the post_test_hook.sh to add that17:01
mriedemidk, like dhellmann said, this is really just pain for the developers that are doing the CI17:03
mriedemi'm not really sure how you'd advertise the gnocchi version that ceilometer stable/liberty should be using17:03
gordcmriedem: oh. yeah, we didn't have gnocchiclient released in liberty... or it was very late17:03
mriedemsince it's not in requirements anywhere17:03
mriedemit seems like gnocchi should live in stable/liberty upper-constraints.txt17:04
mriedemand would be ===1.317:04
mriedemhowever, i don't think we have anything that would be testing with that17:04
gordcmriedem: so i just want to clarify, gnocchiclient doesn't actually exists in liberty timeframe17:05
mriedemyeah i know17:06
gordcso we want gnocchi in requirements for stable/liberty and gnocchiclient in requirements post-liberty?17:08
jd__I don't think we need it in requirement, just pip install from devstack?17:09
jd__from the devstack plugin of Ceilometer I mean17:09
jd__e.g. like we do for some other backends17:10
gordcjd__: we would still have OVERRIDE condition for gnocchi though... like when it runs from gnocchi pov17:10
gordci was under impression people didn't like using OVERRIDE period.17:11
gordci guess start by moving to pypi? maybe let the liberty patches merge in meantime.17:12
mriedemi was thinking pip installing gnocchi from the devstack plugin17:13
mriedemwhich it'd be nice if that used upper-constraints from stable/liberty17:13
mriedemb/c you could set that to 1.3 in there17:14
mriedemand upper-constraints can be used as the manifest for what we've tested against for stable/liberty17:14
mriedemwhich is what we're telling users17:14
jd__we can pip install gnocchi from the Ceilometer plugin _except_ if Gnocchi is already installed – which would be the case if it's pulled by devstack when running the job from Gnocchi pov, WDYT gordc ?17:14
gordcmriedem: we want to add gnocchi to global-reqs only for stable/liberty? is there something we do for elasticsearch in liberty? we don't actually dictate versioning for that do we?17:18
mriedemgordc: not global-reqs since that's really only used for syncing to project's requirements files17:19
mriedembut upper-constraints should be used when pip installing things in devstack17:19
mriedemi think anyway17:19
gordcmriedem: lol i like the confidence. :)17:19
mriedemwell i didn't work on the mechanics17:20
mriedemthat's more lifeless i think17:20
mriedemso maybe we should just table this until lifeless is around and see what he thinks17:20
gordcmriedem: so simliar to what jd__  is suggesting?17:20
jd__we can pip install without going to upper req, just "pip install gnocchi>=1.3,<1.4" directly from the plugin if that's what we support for example17:21
gordcmriedem: we can try to do pip install and then use devstack when check runs in gnocchi gate... i just found a lot of random gotchas of different places setting/ignoring different things17:21
openstackgerritMerged openstack/releases: Fix week starting date  https://review.openstack.org/26602317:24
*** dtantsur|brb is now known as dtantsur|afk17:27
dhellmanngordc, jd__, mriedem : sorry I missed the discussion. It sounds like there's a plan though?17:31
gordcdhellmann: somewhat. i think the idea is to explore how to set ceilometer/aodh/heat to use gnocchi through pypi (and see if upper-constraint does anything if it's not in global-reqs)... and gnocchi gate to run through devstack.17:35
dhellmanngordc : ok. constraints do work for things not in g-r -- we use them for secondary dependencies, too, that's sort of the point.17:35
gordcmriedem: same understanding?17:35
gordcdhellmann: cool. i guess that's one step forward.17:36
lifelessmriedem: morning17:50
mriedemi'm just about to grab lunch17:51
lifelessk, ping me later17:51
mriedembackscroll basically says it all - we were just trying to sort out how ceilometer should depend on gnocchi (which doesn't have stable/liberty branch scheme),17:53
mriedemand how to communicate what versions of gnocchi works with stable/liberty ceilometer17:54
openstackgerritMerged openstack/releases: ceilometerclient 2.2.0  https://review.openstack.org/26643818:33
*** mriedem has quit IRC18:40
openstackgerritDoug Hellmann proposed openstack-infra/release-tools: stop sending folks to launchpad milestones for release details  https://review.openstack.org/26606418:43
*** mriedem has joined #openstack-release18:43
dhellmannbswartz : hey, are you still interested in this manilaclient release? https://review.openstack.org/#/c/259799/18:44
stevemardhellmann: re: https://review.openstack.org/#/c/266433/ i thought we didn't change requirements for stable releases?18:50
lifelessmriedem: upper-constraints should apply normally18:50
lifelessmriedem: the most likely issue is that a gnocchi raises a dep to a min version incompatible w/liberty in some way (e.g. oslo.config breaks compat, gnocchi needs latest oslo.config, boom)18:51
mriedemlifeless: yeah so i was thinking gnocchi===1.3 in stable/liberty upper-constraints18:53
mriedemceilometer devstack plugin would install from that18:53
dhellmannstevemar : there's an upper constraints file, and that needs to be updated since this is one of our libs18:54
lifelessit makes sense to me that ceilometer would bring in it's deps ;)18:54
sdaguelifeless: is there any bit issue if nova team makes our normal tox targets all force constraints?18:56
stevemardhellmann: gotcha, i see that now18:57
dhellmannstevemar : sorry, I should have been more detailed in my comment18:58
stevemardhellmann: yeah, wait a tick, those in u-c are the mitaka releases18:59
dhellmannstevemar : ah, ok, let's leave it then. mriedem and some of us are working through how to revert those updates18:59
dhellmannstevemar : I'll go ahead and cut the release19:00
stevemardhellmann: i'll have mitaka ones tomorrow19:03
dhellmannstevemar : ok19:03
bswartzdhellmann: thanks for the ping19:04
bswartzdhellmann: which change merges first? the requirements change or the release?19:05
dhellmannbswartz : we have to do the release before you can use it, but we want the requirements change up for review so we can approve it quickly after the release is done19:05
lifelesssdague: so, we didn't want to have different semantic for e.g. 'py27'19:06
lifelesssdague: since the behaviour of py27 is in the project testing interface19:06
openstackgerritMerged openstack/releases: release keystoneauth and keystoneclient for stable/liberty  https://review.openstack.org/26643319:06
lifelesssdague: there's a technical issue, which is getting the file on disk during CI, which is only wired up for the -constraints jobs19:07
lifelesssdague: and there is the social issue of confusion / inconsistency19:07
lifelesssdague: the former is solvable, if nova cares enough; the later I'm not sure about19:07
bswartzdhellmann: https://review.openstack.org/#/c/266534/19:13
dhellmannbswartz : ack, I'll look at the rest of that review now19:14
sdagueright, there is more concern that you've got to teach everyone a new command to run tests19:15
sdagueafter we spent 4 years teaching them the old command19:15
dhellmanngordc : the ceilometerclient requirements bump is failing, do you want to debug that? https://review.openstack.org/#/c/266440/19:20
gordcdhellmann: kk. thanks for the heads up19:20
*** yarkot has joined #openstack-release19:22
gordcdhellmann: just hitting recheck so it picks up new ceilometerclient19:23
dhellmanngordc : sounds good, thanks19:24
gordcdhellmann: if i create a depends-on against that upper-constraint patch, would it work?19:25
*** yarkot has quit IRC19:25
gordcwant to see if i can run it against full suite of tests (not just tempest)19:25
dhellmanngordc : a depends-on where? ceilometer?19:25
gordcright. or from aodh.. that links to requirements bump.19:26
gordcsomething that*19:26
dhellmannthat would work for the d-g jobs, but not unit tests19:26
dhellmannbecause of the way requirements syncing works19:27
gordckk. i'll give that a try as well.19:27
lifelesssdague: so folk on master should be fine most of the time running unconstrained19:58
lifelesssdague: I get there is an education aspect - I'm not trying to argue we've got the competing balances right19:59
sdaguelifeless: I think that confusion of "fails for me" but works upstream is something we want to avoid20:00
sdaguealso, it appears tox -epy27-constraints doesn't work for me locally20:01
sdagueis there something else required for it20:01
lifelesssdague: is it passing in CI ?20:30
lifelesssdague: define 'doesn't work' ?20:30
gordclifeless: some additional info, we should think of gnocchi == mysql. in liberty, ceilometer doesn't have req on gnocchiclient since it doesn't exist yet. in mitaka, ceilometer has req on gnocchiclient but it's defined in test-requirements becuase it's a driver (so not a hard requirement for ceilometer itself)20:30
sdagueI am trying to run it on my desktop20:30
sdaguegit clone nova; tox -e py27-constraints20:31
*** topolznc has joined #openstack-release20:35
*** topolznc is now known as topol20:52
*** topol is now known as Guest4533520:52
*** Guest45335 is now known as topol_20:57
*** topol_ is now known as Guest1258120:58
*** Guest12581 is now known as topol220:58
*** LouisF has joined #openstack-release20:58
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Oslo Releases for week of Jan 11, 2016  https://review.openstack.org/26567321:05
dimsdhellmann : waiting for your review :) ^^^21:06
dhellmanndims : ack, did you address josh's comments?21:06
*** topol2 is now known as topol_21:06
dimsdhellmann : yep21:07
dhellmanndims : ok21:07
*** topol_ is now known as topol21:07
sdaguelifeless: this is what I get when I try to run py27-contraints locally - http://paste.openstack.org/show/483663/21:08
dhellmannsdague : you have the latest version of pip?21:10
dhellmannI wonder if tox is hiding a better error in one of its log files21:10
dhellmann"invocation error" doesn't mean much21:10
dhellmannnor does "Missing distribution spec" I guess21:11
dhellmannI wonder if the file has a blank line or something21:11
*** sigmavirus24 is now known as sigmavirus24_awa21:12
*** sigmavirus24_awa is now known as sigmavirus2421:13
sdagueI don't know21:13
openstackgerritMerged openstack/releases: Release python-manilaclient 1.6  https://review.openstack.org/25979921:17
ttxWow zuul is having a hard time today21:20
openstackgerritJoshua Harlow proposed openstack/releases: Oslo Releases for week of Jan 11, 2016  https://review.openstack.org/26567321:21
dimsttx: clarkb mentioned that nodepool is busted21:21
lifelessgordc: we're moving to having drivers exposed as extras21:51
lifelessgordc: so you'd say ceilometer[gnocchi]21:52
lifelesssdague: thats weird21:53
lifelessblank lines should be ok21:53
lifelesssdague: on master or stable/liberty?21:53
gordclifeless: yep, we have that in gnocchi and aodh :)21:53
gordcjust haven't gotten aroudn to it for ceilometer21:53
lifelesssdague: ah, I see. let me try locally21:53
lifelesssdague: the {opts} thing is suspicious, since thats the line that its trying to parse21:54
lifelessbut IIRC thats a tox substitution variable21:54
lifelesssdague: '--force-reinstall', '{opts}',21:54
lifelesssdague: line 721:55
lifelesssdague: I suspect a too-old tox? What is your tox version ?21:55
openstackgerritDavanum Srinivas (dims) proposed openstack/releases: Fix pep8 errors  https://review.openstack.org/26659422:11
dimslifeless : dhellmann : looks like something changed and we have pep8 errors in releases/ repo ^^22:12
*** dims has quit IRC22:15
*** gordc has quit IRC22:32
LouisFdhellmann: i want to get networking-sfc released to mitaka22:44
LouisFdhellmann: do i need to add networking-sfc.yaml to deliverables.mitaka?22:45
*** mriedem has quit IRC22:54
*** sigmavirus24 is now known as sigmavirus24_awa23:01
*** Kiall has quit IRC23:03
*** Kiall has joined #openstack-release23:04
*** sigmavirus24_awa is now known as sigmavirus2423:38
*** sigmavirus24 is now known as sigmavirus24_awa23:47
*** sigmavirus24_awa is now known as sigmavirus2423:48

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!