*** lbragstad has quit IRC | 01:27 | |
openstackgerrit | Paul Belanger proposed openstack/requirements master: [WIP] test wheel builds on bionic https://review.openstack.org/550254 | 01:40 |
---|---|---|
openstackgerrit | Paul Belanger proposed openstack/requirements master: [WIP] test wheel builds on bionic https://review.openstack.org/550254 | 01:42 |
openstackgerrit | Paul Belanger proposed openstack/requirements master: Begin testing wheel builds with ubuntu-bionic https://review.openstack.org/550253 | 01:45 |
openstackgerrit | Paul Belanger proposed openstack/requirements master: [WIP] test wheel builds on bionic https://review.openstack.org/550254 | 01:45 |
*** edmondsw has quit IRC | 02:21 | |
*** edmondsw has joined #openstack-requirements | 02:22 | |
*** edmondsw has quit IRC | 02:26 | |
*** andreas_s has joined #openstack-requirements | 02:43 | |
*** andreas_s has quit IRC | 02:48 | |
openstackgerrit | Paul Belanger proposed openstack/requirements master: Update bindep.txt for ubuntu-bionic https://review.openstack.org/550254 | 03:04 |
*** edmondsw has joined #openstack-requirements | 05:09 | |
*** udesale has joined #openstack-requirements | 05:10 | |
*** edmondsw has quit IRC | 05:14 | |
*** prometheanfire has quit IRC | 06:56 | |
*** edmondsw has joined #openstack-requirements | 06:58 | |
*** coolsvap has joined #openstack-requirements | 07:00 | |
tonyb | Are we doing a thing? | 07:02 |
*** edmondsw has quit IRC | 07:03 | |
coolsvap | I dont see prometheanfire | 07:03 |
dirk | o/ | 07:03 |
dirk | maybe he fell asleep | 07:04 |
tonyb | #startmeeting requirements | 07:04 |
openstack | Meeting started Wed Mar 7 07:04:09 2018 UTC and is due to finish in 60 minutes. The chair is tonyb. Information about MeetBot at http://wiki.debian.org/MeetBot. | 07:04 |
openstack | Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. | 07:04 |
*** openstack changes topic to " (Meeting topic: requirements)" | 07:04 | |
openstack | The meeting name has been set to 'requirements' | 07:04 |
dirk | o/ | 07:04 |
tonyb | The queue .... https://review.openstack.org/#/q/status:open+project:openstack/requirements | 07:04 |
tonyb | has anyone looked at it in the last 48 hours? | 07:05 |
coolsvap | o/ | 07:05 |
* coolsvap looking at it now | 07:05 | |
dirk | a bit yeah, I don't think there is anything controversial atm | 07:06 |
dirk | actually it seems people are still on PTG-travel or other post activitie | 07:06 |
tonyb | The good news is tripleo has branched, so we just need to verify that kolla and OSA have and then we can really-really thaw | 07:06 |
dirk | the interesting thing is the switch to py36 testing (add bionic support) | 07:06 |
dirk | tripleo was the last project | 07:07 |
dirk | and we already started merging uc over the weekend | 07:07 |
tonyb | dirk: Yeah, we have easy access to wheel builders now and can add similat for tox-py36 when we're ready | 07:07 |
tonyb | Oh cool | 07:07 |
* tonyb will fix up constraints generation and add py36 tomorrow if y'all don't beat me to it | 07:08 | |
dirk | #action tonyb fix up constraints generation and add py36 tomorrow | 07:09 |
dirk | happy to defer it to you :-) | 07:09 |
* dirk is pretty busy today | 07:09 | |
tonyb | ;P | 07:10 |
dirk | I might look at adding tumbleweed testing somewhere, but I don't see a *lot* of value in there | 07:10 |
*** pfire has joined #openstack-requirements | 07:10 | |
pfire | internet is down | 07:10 |
coolsvap | pfire: o/ | 07:10 |
tonyb | dirk: Havign new stuff would be good but it would mostly be a duplicate job | 07:11 |
dirk | or asking it differently: would it make sense to run a devstack-integration on more than ubuntu on constraints changes? I don't see a good reason atm | 07:11 |
tonyb | dirk: if we were using distro packages then it make a lot of sense but as we're just pulling most stuff from pypi we don't gain a lot IMO | 07:11 |
dirk | tonyb: agreed | 07:12 |
tonyb | #topic ptg | 07:13 |
*** openstack changes topic to "ptg (Meeting topic: requirements)" | 07:13 | |
dirk | one thing I haven't been following - does anyone want to look at getting rid of outdated/unused deps? | 07:13 |
tonyb | Anything we need to follow up on from the PTG | 07:13 |
pfire | talked to Doug earlier today | 07:14 |
tonyb | pfire: Yeah I read it | 07:15 |
dirk | yeah, there was some proposal from doug, I haven't fully read the backlog yet | 07:15 |
pfire | I think we are on close to the same page... | 07:15 |
tonyb | I think we're going to struggle to get everyon in the same discussion :( | 07:15 |
pfire | ya | 07:15 |
pfire | if you can talk to him don't wait for me | 07:16 |
tonyb | pfire: I'm not 100% certain I agree with that, and dhellmann's proposal will make the lower-constraints work harder | 07:16 |
tonyb | I get thet dhellmann isn't signing up for oding that which is fine but we still want to do it | 07:16 |
dirk | maybe all that we need to agree on is the first step, as thats the one we've been not doing so far | 07:16 |
pfire | by decoupling the carrot and stick? | 07:17 |
tonyb | ? | 07:17 |
pfire | carrot, per project reqs | 07:17 |
pfire | stick lc testing | 07:17 |
tonyb | I still don't follow | 07:17 |
pfire | we can talk later, when I'm not on the phone | 07:18 |
tonyb | okay | 07:18 |
tonyb | Anthing else from the PTG? | 07:19 |
dirk | was there any discussion on the pycrypto thing? | 07:19 |
dirk | or is the transition to cryptography settled? | 07:20 |
pfire | just to remind them more often, offering to help when needed | 07:20 |
tonyb | dirk: No I think we need to push for that to be an 'S' goal | 07:21 |
pfire | also pycrypyodome is the emergency fallback, kinda | 07:21 |
*** andreas_s has joined #openstack-requirements | 07:21 | |
pfire | why s? | 07:21 |
tonyb | pfire: becaus the goals for R are settled already aren't they? | 07:22 |
dirk | earlier would be better though | 07:22 |
dirk | it started to hurt about last cycle | 07:22 |
dirk | I don't think its that big of a mess to be an official goal | 07:23 |
pfire | yes, though I started bugging people about it before the ptg | 07:23 |
tonyb | dirk: Sure, we can work on it without a goal and it can be done ad-hoc but if it's still an issue in 4 months we shoudl push for a goal to get the work complete | 07:23 |
dirk | but I guess dependency-cleanliness is not a priority anymore with containerization | 07:23 |
pfire | tony: cross project goal or reqs goal? | 07:24 |
tonyb | pfire: community wide goal | 07:24 |
tonyb | I agree it's apoor candidate but it's the only tool we have so .... | 07:25 |
pfire | ok, didn't know that was needed | 07:25 |
dirk | tonyb: wfm. Or maybe draft a generic goal of reducing overlapping dependencies | 07:25 |
pfire | ya | 07:25 |
tonyb | We could perhaps aim for a couple of items on our problem list | 07:25 |
tonyb | Sure | 07:25 |
pfire | it does also fall under that overlap | 07:25 |
tonyb | I just think we've been asking for 2 years now so we need to try somethign else | 07:26 |
pfire | I agree, cross project seems like a good idea | 07:26 |
dirk | I tried to look at deps that haven't released a new version for 2+ years | 07:26 |
dirk | As those might not be a good dependency | 07:27 |
dirk | That gives quite a list of candidate | 07:27 |
tonyb | dirk: Wow that's a big chunk of work. ;P | 07:27 |
dirk | Well, we don't need to eliminate all of them | 07:28 |
dirk | But the problematic ones | 07:28 |
pfire | for now I think we should limit our focus, but that's good for next cycle | 07:28 |
dirk | A crypto library without updates sounds smelly | 07:28 |
tonyb | sure | 07:29 |
tonyb | okay aything else? | 07:30 |
tonyb | #topic open discussion | 07:31 |
*** openstack changes topic to "open discussion (Meeting topic: requirements)" | 07:31 | |
tonyb | coolsvap: which city are you in again? | 07:31 |
pfire | ok, gonna leave for now then | 07:31 |
pfire | cell phone and all | 07:31 |
tonyb | pfire: okay | 07:32 |
*** pfire has quit IRC | 07:32 | |
dirk | tonyb: the other obvious candidate for dropping is tempest-lib | 07:32 |
coolsvap | tonyb: Pune India | 07:33 |
dirk | ah we moved on already, sorry | 07:33 |
tonyb | dirk: Yeah I guess if tempest is the right thin to use there we shoudl do that | 07:33 |
dirk | well, any change to tempest-lib is -2'ed so its unmaintainable per excellence :-) but still there seems to be things depending on it | 07:33 |
tonyb | dirk: no biggie the topic is very fluid | 07:34 |
tonyb | dirk: Okay I'll reach out to thr QA team for help on that one | 07:34 |
tonyb | .... So meeting times, its the bi-annual veryone moves clock thing over the next few weeks | 07:34 |
tonyb | when that's done this meeing looks like: https://www.worldtimebuddy.com/?qm=1&lid=2147714,12,6,1259229&h=2147714&date=2018-4-4&sln=17.5-18 | 07:35 |
openstackgerrit | Dirk Mueller proposed openstack/requirements master: Drop tempest-lib from g-r https://review.openstack.org/550375 | 07:35 |
tonyb | which is a bit silly for prometheanfire, dhellman and dims | 07:35 |
tonyb | so we shoudl probably look at what times are do able for us as a group | 07:36 |
dirk | yeah, its getting late for prometheanfire | 07:36 |
tonyb | ... Or give up and switch to alternatine meetins and rely on in channel comms to keep us in sync | 07:36 |
tonyb | thoughts? | 07:37 |
dirk | tonyb: the question is if you want to do a meeting at 11pm+ :-) | 07:37 |
tonyb | dirk: I do not ;P | 07:37 |
tonyb | https://www.worldtimebuddy.com/?qm=1&lid=2147714,12,6,1259229&h=2147714&date=2018-4-4&sln=21.5-22.5 isn't too bad | 07:38 |
tonyb | but I don't think prometheafire is a early riser | 07:38 |
dirk | 630 am might be tough for pfire | 07:38 |
tonyb | anyway start thinking and and we can disccuss it in channel or on the mailing list | 07:38 |
* tonyb really dislikes moving meetings every 6 months | 07:39 | |
dirk | so I think we used to have the slot at noon | 07:39 |
dirk | e.g. 8pm sydney 5am central | 07:39 |
dirk | but at this point its mostly prometheanfirew who needs to decide when he wants to be awake :-) | 07:40 |
tonyb | Yup | 07:40 |
dirk | the alternative is something like 6am european | 07:40 |
dirk | which is however still 11pm central - so probably not good for dims etc | 07:41 |
tonyb | Yeah. | 07:41 |
tonyb | It's a massive pain :( | 07:41 |
dirk | tonyb: so 9pm would be good for you? | 07:42 |
dirk | 6am central might be acceptable for some | 07:42 |
tonyb | Anyway I think we shoudl endmeeting as we're not going to decide now but we have a starting point for conversation | 07:42 |
dirk | agreed | 07:42 |
tonyb | dirk: good? no, but I can do it ;P | 07:42 |
tonyb | Thanks everyone | 07:43 |
tonyb | #endmeeting | 07:43 |
*** openstack changes topic to "OpenStack Requirements - IRC meetngs on Wednesdays @ 07:00 UTC in here in #openstack-requirements - See agenda @ http://tinyurl.com/h44ryuw - IRC channel is *LOGGED* @ http://tinyurl.com/j38rk24" | 07:43 | |
openstack | Meeting ended Wed Mar 7 07:43:40 2018 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4) | 07:43 |
openstack | Minutes: http://eavesdrop.openstack.org/meetings/requirements/2018/requirements.2018-03-07-07.04.html | 07:43 |
openstack | Minutes (text): http://eavesdrop.openstack.org/meetings/requirements/2018/requirements.2018-03-07-07.04.txt | 07:43 |
openstack | Log: http://eavesdrop.openstack.org/meetings/requirements/2018/requirements.2018-03-07-07.04.log.html | 07:43 |
dirk | I'd be willing to experiment with asynchronous meeting tbh | 07:44 |
dirk | with some discipline I have made good experiences with an etherpad-discussion over the course of a day | 07:45 |
dirk | because things can be commented on in place, but without the need for time synchroniticity | 07:45 |
dirk | as a food for thought ;-) | 07:46 |
tonyb | dirk: Yeah | 07:53 |
*** florianf has joined #openstack-requirements | 08:41 | |
openstackgerrit | Merged openstack/requirements master: update constraint for openstackdocstheme to new release 1.19.0 https://review.openstack.org/549998 | 08:44 |
openstackgerrit | Merged openstack/requirements master: update constraint for openstacksdk to new release 0.12.0 https://review.openstack.org/549983 | 08:44 |
openstackgerrit | Merged openstack/requirements master: Adds castellan-ui https://review.openstack.org/550262 | 08:44 |
openstackgerrit | Merged openstack/requirements master: update monasca-common min 2.7.0 https://review.openstack.org/543415 | 08:44 |
openstackgerrit | Merged openstack/requirements master: Begin testing wheel builds with ubuntu-bionic https://review.openstack.org/550253 | 08:44 |
*** edmondsw has joined #openstack-requirements | 08:46 | |
*** ralonsoh has joined #openstack-requirements | 08:48 | |
*** ralonsoh_ has joined #openstack-requirements | 08:48 | |
*** edmondsw has quit IRC | 08:50 | |
openstackgerrit | Merged openstack/requirements master: Update bindep.txt for ubuntu-bionic https://review.openstack.org/550254 | 09:10 |
openstackgerrit | Merged openstack/requirements master: Add documentation for lower-constraints.txt https://review.openstack.org/549593 | 09:26 |
openstackgerrit | Dirk Mueller proposed openstack/requirements master: Drop tempest-lib from g-r https://review.openstack.org/550375 | 09:52 |
*** ralonsoh_ has quit IRC | 10:17 | |
*** edmondsw has joined #openstack-requirements | 10:34 | |
*** edmondsw has quit IRC | 10:38 | |
*** coolsvap has quit IRC | 10:49 | |
*** udesale_ has joined #openstack-requirements | 11:19 | |
*** udesale has quit IRC | 11:21 | |
*** udesale has joined #openstack-requirements | 11:21 | |
*** udesale_ has quit IRC | 11:24 | |
*** udesale has quit IRC | 11:36 | |
*** vpickard_ is now known as vpickard | 12:19 | |
*** mattoliverau has quit IRC | 12:21 | |
*** mattoliverau has joined #openstack-requirements | 12:22 | |
*** udesale has joined #openstack-requirements | 12:39 | |
*** lbragstad has joined #openstack-requirements | 12:39 | |
*** ameeda has joined #openstack-requirements | 12:59 | |
*** ameeda has left #openstack-requirements | 13:13 | |
*** prometheanfire has joined #openstack-requirements | 13:20 | |
*** edmondsw_ has joined #openstack-requirements | 13:25 | |
openstackgerrit | Monty Taylor proposed openstack/requirements master: Add neutron and horizon to global requirements https://review.openstack.org/550475 | 13:55 |
*** lbragstad has quit IRC | 14:15 | |
*** lbragstad has joined #openstack-requirements | 14:21 | |
*** pabelanger has quit IRC | 15:04 | |
*** migi has quit IRC | 15:04 | |
*** edmondsw_ is now known as edmondsw | 15:13 | |
*** pabelanger has joined #openstack-requirements | 15:18 | |
*** udesale has quit IRC | 15:45 | |
openstackgerrit | Merged openstack/requirements master: Limit construct lib to before 2.9 https://review.openstack.org/549866 | 15:46 |
*** andreas_s has quit IRC | 16:14 | |
*** andreas_s has joined #openstack-requirements | 16:19 | |
*** andreas_s has quit IRC | 16:28 | |
*** andreas_s has joined #openstack-requirements | 16:33 | |
*** andreas_s has quit IRC | 16:47 | |
*** andreas_s has joined #openstack-requirements | 16:49 | |
*** andreas_s has quit IRC | 16:58 | |
*** florianf has quit IRC | 17:21 | |
*** andreas_s has joined #openstack-requirements | 17:22 | |
*** andreas_s has quit IRC | 17:41 | |
*** andreas_s has joined #openstack-requirements | 17:44 | |
*** andreas_s has quit IRC | 17:57 | |
*** andreas_s has joined #openstack-requirements | 18:02 | |
*** ralonsoh has quit IRC | 18:14 | |
*** andreas_s has quit IRC | 18:16 | |
dhellmann | dirk : ++ | 18:20 |
dhellmann | (for more async communication) | 18:21 |
*** andreas_s has joined #openstack-requirements | 18:22 | |
*** andreas_s has quit IRC | 18:36 | |
*** andreas_s has joined #openstack-requirements | 18:41 | |
*** sdake has joined #openstack-requirements | 18:59 | |
*** sdake has quit IRC | 18:59 | |
*** sdake has joined #openstack-requirements | 18:59 | |
*** andreas_s has quit IRC | 18:59 | |
prometheanfire | ugh, looks like gentoo was still using the old msgpack | 19:34 |
openstackgerrit | Emmet Hikory proposed openstack/requirements master: Begin testing wheel builds with Ubuntu Xenial ARM64 https://review.openstack.org/550578 | 19:40 |
pabelanger | is there a plan to delete stable/newton branch on requirements? I only ask because we are still building wheels, and unsure if jobs are still using that branch | 19:46 |
openstackgerrit | Emmet Hikory proposed openstack/requirements master: Begin testing wheel builds with Ubuntu Xenial ARM64 https://review.openstack.org/550578 | 19:48 |
prometheanfire | pabelanger: no | 20:28 |
dhellmann | prometheanfire, dirk, tonyb : do you all feel that the notes I drew up yesterday are ready to be turned into a plan and published on the -dev list for feedback? | 20:42 |
prometheanfire | I'd like a final review of the email, but generally yes I think | 20:43 |
prometheanfire | tony had problems but I don't know what they were (I was offline due to cable company) | 20:44 |
dhellmann | prometheanfire : oh, well, the idea is I would write the email as a "this is what I plan to do" thing and you'd comment there and we'd come to agreement. We don't have a specs repo, or I'd just use that. | 20:47 |
prometheanfire | I'd like to at least wait til tony acks | 20:48 |
dhellmann | sure | 20:49 |
dhellmann | is tonyb's comment the one in purple on line 88? | 20:50 |
dhellmann | oh, no, I see it in the meeting log scrollback | 20:51 |
prometheanfire | I don't have that, I should look at irclogs | 20:51 |
dhellmann | tonyb : I'm not sure I see how the lower constraints work becomes harder. That has to be per-repo anyway, so whatever tool we have to produce the list would need to work with requirements files input from a repo and produce the lower constraints. | 20:52 |
prometheanfire | tonyb: ya, that time isn't too bad (around noon UTC) | 20:52 |
dhellmann | I guess the global list was going to fill in minimum values for things we don't have listed explicitly in a given requirements file, but we can always get the old version of the file to do that. Or we could delay removing the minimum settings. | 20:52 |
prometheanfire | dhellmann: removing minimum settings was on the table? | 20:54 |
dhellmann | I suppose one way to do that would be to build a constraints file using the minimum values from the g-r list, then install the packages in the requirements files, then run pip freeze to see what packages were actually installed and use that list as the starting point for lower-constraints.txt for that repo. | 20:54 |
dhellmann | prometheanfire : yes, step 3 on line 113 | 20:54 |
prometheanfire | line 113 of? | 20:54 |
dhellmann | we don't want to imply that those global minimums are correct | 20:54 |
dhellmann | https://etherpad.openstack.org/p/ocata-requirements-notes | 20:54 |
prometheanfire | ah | 20:55 |
prometheanfire | in the last cycle we went through and added the mins so that we could do min based testing | 20:55 |
prometheanfire | you'll probably see dirks name a bunch in git blame for that | 20:55 |
dhellmann | we've always had mins in g-r. maybe not for every single line? | 20:55 |
prometheanfire | ya, wasn't for every single one | 20:56 |
dhellmann | ok | 20:56 |
prometheanfire | it is now and is enforced | 20:56 |
dhellmann | I guess with that existing lower-constraints.txt file we can just use that to build a per-project constraints file | 20:57 |
prometheanfire | yep | 20:57 |
prometheanfire | that was the idea | 20:57 |
dhellmann | and what's the plan for defining the test job? python 3 only I hope? and in the repo via the local zuul config? | 21:00 |
prometheanfire | well, we'd like to see at least py3, but given swift is the poster boy... | 21:01 |
prometheanfire | the plan was for at least unit tests to be run with the project's lower constraints | 21:01 |
dhellmann | prometheanfire : like this? https://review.openstack.org/550603 | 21:04 |
prometheanfire | ya, looks about right | 21:06 |
dhellmann | how many repos do we think this is going to apply to? | 21:08 |
prometheanfire | it's opt-in, so whoever wanted to be independent | 21:09 |
dhellmann | "independent"? | 21:09 |
prometheanfire | per-project | 21:09 |
prometheanfire | wanted to stop syncing | 21:09 |
dhellmann | oh. Well, I think we want all projects to stop syncing. Why would we not? | 21:09 |
prometheanfire | then all projects I guess, would be best case | 21:10 |
dhellmann | ok | 21:10 |
dhellmann | that's a lot of patches | 21:10 |
prometheanfire | it is | 21:10 |
dhellmann | I suppose it lets us define the job one time, though | 21:10 |
dhellmann | instead of in each repo | 21:10 |
dhellmann | although that requires us to land all of the tox settings changes before we can turn the job on | 21:11 |
prometheanfire | how do you mean? you mean in project-config? | 21:11 |
dhellmann | yeah | 21:11 |
prometheanfire | ya | 21:11 |
dhellmann | I don't think we want that | 21:11 |
dhellmann | at least not to start | 21:11 |
dhellmann | it would delay getting things working | 21:11 |
dhellmann | yay, the job is running against that patch | 21:12 |
dhellmann | do we need that in the gate queue, too? | 21:12 |
dhellmann | if all the changes we need are in the project repo, it's probably scriptable. The question is whether we actually want all projects to have these tests, I guess. | 21:15 |
*** andreas_s has joined #openstack-requirements | 21:15 | |
dhellmann | I mean, sure, we do. But is it a requirement or is it a desire? | 21:16 |
prometheanfire | the cost of testing should be relatively low, gate allows us to test as it appears after merge, I think | 21:16 |
prometheanfire | if it's not a requirement I'm not sure it's going to happen | 21:16 |
dhellmann | yeah, I'm more worried about adding 200+ new test definitions than I am about running the extra check and gate job on each patch. | 21:16 |
prometheanfire | you mean the work in getting it merged into all the repos? | 21:17 |
dhellmann | right | 21:17 |
dhellmann | I don't want to block the rest of the syncing changes on that | 21:17 |
dhellmann | projects.txt has 324 lines | 21:17 |
prometheanfire | ya, that part is going to suck, but don't we do the same thing for tox.ini changes? | 21:18 |
dhellmann | for the constraints URL? | 21:18 |
prometheanfire | and we can do it slowly, I get the impression you think it has to all merge at once | 21:18 |
prometheanfire | ya, like the constraints url (stable branch) updates | 21:18 |
dhellmann | well, you had been saying this was a prerequisite for stoping syncing | 21:19 |
dhellmann | those patches are proposed by a bot now | 21:19 |
dhellmann | but yeah, I guess since this looks pretty scriptable I can add it to the work I would be doing | 21:19 |
dhellmann | assuming folks agree with the rest of the general plan and the way I've prepared the sample for oslo.config makes sense | 21:20 |
prometheanfire | the sync could check for the job definition of the lower check/gate and if found skip | 21:20 |
*** andreas_s has quit IRC | 21:20 | |
prometheanfire | we'd still have to do cleanup and remove from projects.txt, but at least then they'd be desync'd | 21:21 |
prometheanfire | I think that's what you are looking for? | 21:21 |
prometheanfire | I think I'm happy with this, would be good to get the others view to be sure | 21:21 |
prometheanfire | we'd like the check/gate to be voting as well, but that should hopefully go without saying :D | 21:22 |
dhellmann | so you're saying change the sync job to skip the repo if it has the lower-constraints tox environment? | 21:23 |
dhellmann | I still don't like coupling these 2 things together | 21:24 |
dhellmann | according to fungi in -infra if we propose too many zuul config changes at one time it will break zuul | 21:24 |
dhellmann | so we have to batch them | 21:24 |
prometheanfire | that's fine | 21:24 |
prometheanfire | how do you mean coupling them together? | 21:24 |
dhellmann | adding the l-c test and stopping the sync job | 21:25 |
fungi | concurrent configuration changes running tests | 21:25 |
prometheanfire | oh, ya | 21:25 |
prometheanfire | fungi: what's a good number? | 21:25 |
dhellmann | we thought ~10 at a time would be safe | 21:26 |
fungi | 10 at a time is probably safe if you let most of them finish testing before you push more | 21:26 |
dhellmann | oh! they can be up for review, just not running tests all at once? | 21:26 |
dhellmann | well, that's a bit better | 21:26 |
dhellmann | I figured we needed to land the batch before proposing more | 21:26 |
fungi | for every proposed zuul configuration change, it assembles a speculative future configuration state and our configuration is pretty huge already | 21:26 |
fungi | it needs to hold all those prospective configuration states in memory while it's running jobs against them | 21:27 |
dhellmann | we could define the job in project-config and make it non-voting, but then every project team would have to make it voting with a separate patch to project-config so I'm not sure we gain that much that way | 21:28 |
dhellmann | way more review bandwidth needed from the infra team, for example | 21:28 |
fungi | we've seen pushing ~50 in parallel basically max out the 30gb of ram we have on that system | 21:28 |
prometheanfire | dhellmann: so make it voting by default or...? | 21:29 |
dhellmann | prometheanfire : if we define the job centrally and add it to every repo voting then we have to already have the tox settings in place in all 324 repos before we turn on the job | 21:29 |
dhellmann | if we define it centrally non-voting, then project teams can turn it to be voting as they get it working, but that's potentially 324 more patches for infra to have to review | 21:30 |
prometheanfire | ya | 21:30 |
dhellmann | if we define it in tree, then we don't need infra to review anything and project teams can each review 1 patch per repo | 21:30 |
prometheanfire | I think defining it in tox before we start with project-config | 21:30 |
prometheanfire | the problem with that is that once the job is added stuff will fail til the project teams fix it | 21:31 |
dhellmann | the safest way to do it is to add the job in-tree first; then after they all work we can replace the per-repo job with a centrally defined job | 21:32 |
dhellmann | then remove the per-repo job | 21:32 |
prometheanfire | ya, that would be the safest, a bunch of tracking though, but safer | 21:32 |
dhellmann | I'm extending the etherpad with these notes | 21:33 |
dhellmann | yay, the job passed for oslo.config | 21:35 |
prometheanfire | ya, looks good to me | 21:37 |
prometheanfire | reading from 116 on | 21:38 |
prometheanfire | I don't see an update for the pass yet :P | 21:38 |
dhellmann | 'update for the pass' | 21:40 |
dhellmann | ? | 21:40 |
prometheanfire | the test results listed https://review.openstack.org/#/c/550603 | 21:44 |
dhellmann | oh, it's still running a dsvm job | 21:48 |
dhellmann | see http://zuul.openstack.org | 21:48 |
prometheanfire | ah | 21:49 |
openstackgerrit | Dirk Mueller proposed openstack/requirements master: Drop tempest-lib from g-r https://review.openstack.org/550375 | 21:50 |
dhellmann | ugh, I just realized with all of the variations for how to configure zuul the script to add the job may not be trivial | 21:50 |
dhellmann | lots of projects have .zuul.yaml or .zuul.d or zuul.d | 21:51 |
prometheanfire | ya :| forgot about that | 21:51 |
prometheanfire | dhellmann: I see you are around :P | 21:51 |
dhellmann | we at least have some conventions in place so the latter 2 cases are the same | 21:51 |
prometheanfire | ya | 21:52 |
dhellmann | auto-editing existing yaml files can be messy | 21:53 |
tonyb | dhellmann: In general I think we're on the same page. I need to think through how indepenant that lets projects be and if that's a problem or not. Also while I get that you're not signing up for the lower-constratints work I think that removing minimums from g-r makes that work really hard so I'm inclined to leave it there | 22:10 |
tonyb | dhellmann: which job are we adding? | 22:11 |
dhellmann | https://review.openstack.org/#/c/550603/4 | 22:12 |
dhellmann | tonyb : there is now a lower-constraints.txt file in the requirements repo as well, but as soon as we stop syncing both sets of mins will be lies | 22:12 |
*** openstackstatus has quit IRC | 22:13 | |
*** openstack has joined #openstack-requirements | 22:15 | |
*** ChanServ sets mode: +o openstack | 22:15 | |
prometheanfire | dhellmann: like swift :P | 22:15 |
tonyb | dhellmann: Oh the plan was to work with infra to provide a varient of the tox jobs that would do that for us | 22:15 |
dhellmann | tonyb : do the py35 jobs not run within tox now? I can't remember the state of that. | 22:16 |
dhellmann | but see also the logic in the etherpad about why we don't want to do a central job first | 22:17 |
tonyb | dhellmann: the plan we had in Denver was that the check tool wouldn't allow a project to up a minium above what's in g-r, without first upping it there. In much the same was as today so that mimimum in gor- could be used to regernate the lower-constraints.txt in requirements | 22:17 |
tonyb | dhellmann: No they're runnign in tox. | 22:17 |
dhellmann | tonyb : ok, that violates the whole point of doing this work | 22:18 |
dhellmann | the point is we don't want projects to have to have the same minimums as anyone at all | 22:18 |
prometheanfire | if a project could raise the min, the effective min/max would both be as defined as whatever is in UC | 22:18 |
tonyb | dhellmann: we discussed a series of chnages to make the lower-constratints) gateing a little better | 22:18 |
tonyb | dhellmann: No the same, just we track the highest minimum in g-r | 22:18 |
dhellmann | I don't really see the point to doing that extra work | 22:20 |
prometheanfire | tonyb: so if a project raised their min we'd just raise our min (via a bot that checks all projects?) | 22:20 |
tonyb | $project a can have a no minimum, project b can have >=1.0.0 pand projects{n..m} can have any they like ... with the proviso that the highest minimum across all of openstack is tracked in g-r | 22:20 |
dhellmann | I'm trying to simplify all of this | 22:20 |
tonyb | prometheanfire: Sure we could do it that way but that isn't what we discuessed | 22:20 |
tonyb | prometheanfire: and someone need to write it | 22:20 |
prometheanfire | tonyb: sure, was just thinking how it'd end up working, doesn't have to be that way | 22:20 |
prometheanfire | but automation is nice | 22:21 |
dhellmann | we want projects to express valid minimums. In the absence of "valid" we want something. We don't care that any 2 projects have the same value, because the only time the minimum should ever come into play is if someone is installing a project via pip independently of all other projects and using old versions of packages for some reason. | 22:21 |
prometheanfire | installing without -U | 22:21 |
dhellmann | I want us to stop having bots proposing patches to keep things in sync that we don't need to be in sync. | 22:21 |
tonyb | dhellmann: I agree with all of that. | 22:22 |
prometheanfire | we should ask the question of if we should track mins at all in gr (and != too) | 22:22 |
dhellmann | prometheanfire : I think I've consistently been saying "no" to that question | 22:22 |
prometheanfire | if everything is independant I don't think we need to | 22:23 |
dhellmann | right | 22:23 |
dhellmann | everything is independent | 22:23 |
prometheanfire | dhellmann: you have, and I think I'm coming around | 22:23 |
dhellmann | we need 1 proven set of co-installable versions, and that's our u-c list | 22:23 |
tonyb | dhellmann: as a team we decided that we wanted to be able to do *global* lower constraints, and as a team we decided the only way we could do that was with the approach I described | 22:23 |
prometheanfire | the way we've been thinking of it is that only a few would be independant | 22:23 |
dhellmann | we could add other sets if we want to in the future, but we need to keep at least one for now | 22:23 |
dhellmann | tonyb : ok. I wasn't present for that conversation so I'm sorry for arguing it all again. | 22:24 |
tonyb | dhellmann: I'm not placing a value on *if* we need to do that only sayin that your plan makes that work harder | 22:24 |
dhellmann | I would like to understand why you think that's a useful set of numbers to have | 22:24 |
prometheanfire | I was never a strong 'yes' on tracking global mins other than to help with getting per project stuff started | 22:25 |
tonyb | dhellmann: I'm really happy to revisit the initial premise, and for now I don't consider it a blocker in anyway ... if we can track the miniumums untl we make a call | 22:25 |
prometheanfire | after it's done the only reason would be to allow new projects to be added easilly | 22:25 |
dhellmann | tonyb : that's fine with me. Removing the mins was the last step in the process anyway. | 22:25 |
tonyb | dhellmann: Because some dirstos are very reluctant to up a packaged version of $thing and will only do so if it can be domonstrated as needed | 22:26 |
prometheanfire | gr would stop sending out updates and instead gather them from projects | 22:26 |
dhellmann | prometheanfire : new projects could also assume that their lower bounds are the current upper-bounds, no? | 22:26 |
prometheanfire | it'd be descriptive instead of perscriptive | 22:26 |
prometheanfire | dhellmann: they can, if they are ok with that, sure | 22:26 |
tonyb | this was the lever we came up with to validate it / provide the lever the distros need | 22:26 |
tonyb | In general I'd really like to be able to test the lower bounds to find a rash of bugs we know we have | 22:27 |
tonyb | zigo, used to do that but always too late to be haelpful | 22:27 |
dhellmann | tonyb : I understand that to be true. I also think it's not necessarily the community's responsibility to do the product-level testing for all of the distros by maintaining those lists. | 22:27 |
dhellmann | if we have a $foo-distro-constraints.txt file that someone wants to maintain, I would support variants of test jobs on the platform using those versions. | 22:27 |
dhellmann | but I suspect those distros are already running their own tests using their packages instead, so I'm not sure there's community value in doing the tests upstream. | 22:28 |
dhellmann | testing per-project lower-bounds makes sense to me if we say we want projects to be run on their own, but I think that's a different case from what you're sayimng | 22:29 |
dhellmann | saying | 22:29 |
zigo | tonyb: I never really tested lower bounds, I just happened to use the intersection of what's in Debian stable with what's in requirements. | 22:29 |
zigo | And sometimes, I found issues. | 22:29 |
dhellmann | right, just like we do in rdo | 22:29 |
zigo | dhellmann: BTW, I'm done with packaging Queens, tomorrow, I'm starting functional testing, all with py3. | 22:30 |
dhellmann | zigo : \o/ | 22:30 |
zigo | So far, it went all quite well, appart from manila-ui which is not py3 compliant at all. | 22:30 |
dhellmann | it's time for me to make dinner, so I need to drop off | 22:31 |
prometheanfire | I just upgraded my home cluster to queens, smootest yet | 22:31 |
zigo | (and of course, I'm not talking about swift) | 22:31 |
* prometheanfire knocks on wood | 22:31 | |
zigo | dhellmann: Bon appetit ! :) | 22:31 |
tonyb | dhellmann: rdo tracks u-c aggressively | 22:31 |
prometheanfire | ya, I'm about to relocate | 22:31 |
prometheanfire | OSA tracks it too | 22:31 |
zigo | prometheanfire: Relocate to where? | 22:32 |
prometheanfire | home | 22:32 |
prometheanfire | at work now | 22:32 |
zigo | :P | 22:32 |
prometheanfire | 2018-03-07 22:31:33.313 21152 WARNING oslo_config.cfg [req-b83b8861-7e71-457b-9e5f-a28ee80dc041 - - - - -] Option "auth_uri" from group "keystone_authtoken" is deprecated. Use option "www_authenticate_uri" from group "keystone_authtoken". | 22:32 |
prometheanfire | I'm tired of these values changing all the time btw | 22:32 |
tonyb | dhellmann: I more or less agree with everything you've said | 22:32 |
zigo | yeah. | 22:32 |
zigo | Useless... | 22:32 |
zigo | I also agree with dhellmann about not synching minimums, though I'm afraid I don't trust it's going to work because projects will not do a good job at it. | 22:34 |
tonyb | dhellmann: the question with the per distro constraints file (which we agreed to try but the driver didn't do the work) is how to manage the interxection with packages and pypi ... but that's another issue | 22:34 |
dhellmann | tonyb : yeah, the files would have to list versions on pypi for our normal test jobs to work properly | 22:34 |
zigo | And the minimums will be just wrong. | 22:34 |
dhellmann | zigo : I suspect so, too. jobs like https://review.openstack.org/#/c/550603/4 would help but applying those to all of the repos will be a lot of work. | 22:35 |
prometheanfire | if we gather update into gr.txt (from the set that the projects update/test) instead of push updates from it we'd get what might be valid results | 22:35 |
zigo | It'd be really awesome to have it though ! | 22:36 |
prometheanfire | anyway, i'm walking to my car now | 22:36 |
prometheanfire | be back in a bit | 22:36 |
* zigo goes to sleep | 22:36 | |
tonyb | dhellmann: Yup. | 22:36 |
zigo | 23:30 here | 22:36 |
dhellmann | so I think I'll write up this plan as a long email tomorrow and start circulating it on the -dev list for broader feedback. is that ok with everyone? | 22:36 |
zigo | Got to get up at 6:00 to go to $work ... | 22:36 |
prometheanfire | early | 22:36 |
dhellmann | not as a "this is what we will do" but "this is what we want to do" proposal | 22:37 |
* zigo nods | 22:37 | |
*** edmondsw has quit IRC | 22:38 | |
*** edmondsw has joined #openstack-requirements | 22:39 | |
tonyb | dhellmann: Good plan | 22:41 |
*** openstackstatus has quit IRC | 22:42 | |
*** openstack has joined #openstack-requirements | 22:44 | |
*** ChanServ sets mode: +o openstack | 22:44 | |
*** openstackstatus has joined #openstack-requirements | 22:44 | |
*** ChanServ sets mode: +v openstackstatus | 22:44 | |
*** vpickard is now known as vpickard_ | 23:46 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!