| nibalizer | o/ | 00:00 |
|---|---|---|
| *** richm has quit IRC | 02:32 | |
| EmilienM | clarkb: so I got AssertionError: False is not true : Timed out waiting for 172.24.5.48 to become reachable | 02:53 |
| EmilienM | I guess I need to manage 172.24.5.48 interface | 02:53 |
| EmilienM | sorry, 172.24.5.48 is in fact the floating IP | 02:55 |
| EmilienM | so I'm missing something now | 02:55 |
| EmilienM | pabelanger: the experimental job for unit test is working as expected, I guess we can move forward to having it replacing our current jobs | 03:11 |
| *** asahlin has quit IRC | 03:27 | |
| *** asahlin has joined #openstack-sprint | 03:29 | |
| EmilienM | _ody: I can create vs_bridge with current puppet-vswitch module, in one run | 03:58 |
| *** mrmartin has joined #openstack-sprint | 04:12 | |
| EmilienM | crinkle: I got it work with simple puppet code finally -> https://review.openstack.org/#/c/217352/ | 04:57 |
| *** mrmartin has quit IRC | 05:16 | |
| *** mrmartin has joined #openstack-sprint | 06:45 | |
| *** sbadia has joined #openstack-sprint | 08:48 | |
| EmilienM | good morning sprinters! | 11:49 |
| *** mrmartin has quit IRC | 11:54 | |
| *** rfolco has joined #openstack-sprint | 12:00 | |
| *** cdelatte has quit IRC | 12:00 | |
| *** cdelatte has joined #openstack-sprint | 12:15 | |
| *** cdelatte has quit IRC | 12:41 | |
| *** cdelatte has joined #openstack-sprint | 12:43 | |
| *** mrmartin has joined #openstack-sprint | 12:55 | |
| *** richm has joined #openstack-sprint | 13:23 | |
| *** mestery has joined #openstack-sprint | 13:24 | |
| vinsh_ | meow. | 13:24 |
| *** vinsh_ is now known as Vinsh | 13:24 | |
| *** mestery has quit IRC | 13:26 | |
| *** mestery has joined #openstack-sprint | 13:27 | |
| EmilienM | so pabelanger managed to make the experimental job that runs all unit tests in a single job | 13:31 |
| EmilienM | it's working as expected, now I'm preparing a patch to infra so we move the experimental job to replace others | 13:32 |
| Vinsh | woah, runs all the puppet version one after the other? | 13:32 |
| Vinsh | Is that what you mean by all the unit tests? | 13:32 |
| EmilienM | pabelanger: http://whatis.techtarget.com/definition/six-thinking-hats-retrospective | 13:50 |
| EmilienM | Vinsh: yeah | 13:50 |
| EmilienM | to save nodepool resources | 13:50 |
| Vinsh | heck ya. great | 13:50 |
| pabelanger | Vinsh: it is serial right now, but something could be done to make it parallel | 13:51 |
| *** richm has quit IRC | 13:51 | |
| Vinsh | Still probably not the slowest job of the bunch | 13:52 |
| pabelanger | actually, you could do something like diskimage-builder to get parallel jobs going on a host | 13:52 |
| pabelanger | would have to play with it | 13:52 |
| pabelanger | either way, that patch will save some nodepool resources | 13:52 |
| EmilienM | crinkle, nibalizer: I would like your acks on https://review.openstack.org/#/c/220527/ | 14:52 |
| EmilienM | jamespage: b2 seems promising | 14:54 |
| EmilienM | wow | 14:55 |
| EmilienM | tempest even run, etc | 14:55 |
| jamespage | EmilienM, yeah - I see errors for cinder volumes attachment and boot from volume - but that may be a charm misconfiguration | 14:57 |
| jamespage | all other tempest smokes pass | 14:57 |
| EmilienM | yeah, we are not at this stage yet | 14:57 |
| EmilienM | our functional testing is not testing it | 14:57 |
| EmilienM | awesome | 14:57 |
| EmilienM | jamespage++ | 14:57 |
| EmilienM | jamespage: ironic fails: https://review.openstack.org/#/c/219830/ | 14:58 |
| EmilienM | let me find why | 14:58 |
| EmilienM | jamespage: http://logs.openstack.org/30/219830/2/check/gate-puppet-ironic-puppet-beaker-rspec-dsvm-trusty/6ba23f5/console.html#_2015-09-04_14_28_48_496 | 14:58 |
| jamespage | EmilienM, ironic don't appear to be lining up with everyone else for liberty - https://launchpad.net/ironic/liberty/4.0.0 | 14:59 |
| jamespage | so you still have kilo packaging there - which will be broken atm | 15:00 |
| jamespage | (oh and ironic is also not part of our core testing) | 15:00 |
| EmilienM | anteaya: we would like http://docs.openstack.org/developer/puppet-openstack for now | 15:00 |
| EmilienM | anteaya: I want to double check with crinkle / nibalizer if we could use http://docs.openstack.org/developer/puppet | 15:00 |
| anteaya | okay thanks | 15:00 |
| EmilienM | and collaborate on content | 15:00 |
| EmilienM | that would be the best option | 15:00 |
| EmilienM | but I need their thoughts | 15:00 |
| EmilienM | so for now, http://docs.openstack.org/developer/puppet-openstack | 15:00 |
| EmilienM | it would be a good start | 15:01 |
| EmilienM | and I think we can iterate later | 15:01 |
| anteaya | I will work towards that | 15:01 |
| anteaya | thank you | 15:01 |
| anteaya | well it would be tough to change the name later | 15:02 |
| EmilienM | anteaya: why? | 15:02 |
| anteaya | as the docs would be published to one site according to index engines | 15:02 |
| anteaya | and then switching would require a redirect | 15:02 |
| anteaya | as people would expect them in the place they expect them | 15:03 |
| EmilienM | anteaya: mhh, ok - let's prepare patches now, like if it would be puppet-openstack | 15:03 |
| EmilienM | anteaya: and wait for crinkle / nibalizer to see if we can use 'puppet' | 15:03 |
| anteaya | okay | 15:04 |
| anteaya | best if we agree before the repo is created though | 15:04 |
| anteaya | so thanks | 15:04 |
| EmilienM | anteaya: in the meantime, you can work on the patch though, we will iterate patchsets if needed | 15:04 |
| anteaya | right | 15:06 |
| anteaya | doing so now | 15:06 |
| anteaya | ttp://docs.openstack.org/infra/ | 15:08 |
| *** richm has joined #openstack-sprint | 15:10 | |
| crinkle | EmilienM: I don't really care about the docs name, /puppet and /puppet-openstack both seem fine | 15:17 |
| EmilienM | anteaya: go for /puppet | 15:33 |
| anteaya | okey dokey | 15:34 |
| EmilienM | crinkle: good news, we can bump all liberty to 'proposed' for ubuntu | 15:35 |
| richm | EmilienM: Are you working on https://review.openstack.org/218059? | 15:37 |
| EmilienM | richm: no | 15:37 |
| richm | EmilienM: ok - I will then | 15:37 |
| EmilienM | richm: ack | 15:37 |
| EmilienM | crinkle: last night I changed the code to use puppet in a very simple way on https://review.openstack.org/#/c/217352/ and it worked - today I try to get the jobs green for both ubuntu & centos | 15:41 |
| anteaya | EmilienM crinkle and others, patch to create repo puppet-openstack-docs with publish jobs intended to publish to docs.openstack.org/developer/puppet | 15:46 |
| anteaya | https://review.openstack.org/#/c/220555/4 | 15:46 |
| EmilienM | anteaya: this sounds excellent | 15:48 |
| EmilienM | anteaya: why experimental? | 15:48 |
| anteaya | because it is a new job | 15:48 |
| anteaya | and as a reviewer I really really really like to see new jobs in the experimental pipeline first | 15:48 |
| anteaya | so I need to practice what I preach | 15:49 |
| anteaya | I might need a no-op job then for check | 15:49 |
| anteaya | and gate | 15:49 |
| EmilienM | anteaya: yeah, but that means we need to trigger it manually, right? | 15:53 |
| anteaya | yes | 15:56 |
| EmilienM | since the job is not affecting other projects, I suggest we go directly in check and make the first commit pass CI | 15:57 |
| EmilienM | instead of relying on experimental pipeline that requires manual commands to trigger the job | 15:57 |
| anteaya | okay we can do that | 16:00 |
| anteaya | can you put that comment on the patch? | 16:01 |
| EmilienM | anteaya: sure thing | 16:01 |
| EmilienM | anteaya: done | 16:02 |
| crinkle | EmilienM: what happened to clarkb's suggestion of using the devstack-gate function? | 16:03 |
| anteaya | EmilienM: thanks | 16:04 |
| anteaya | crinkle: are you talking about the docs patch? | 16:04 |
| EmilienM | no | 16:04 |
| EmilienM | crinkle: I could not make it work last night, I got tempest issues when not managing br-ex and also the function failed to run in our CI | 16:04 |
| * anteaya moves on | 16:04 | |
| crinkle | anteaya: no, https://review.openstack.org/#/c/217352 | 16:05 |
| EmilienM | it seems like the function is really working for devstack | 16:05 |
| EmilienM | and I had troubles to debug it | 16:05 |
| anteaya | crinkle: k, thanks | 16:05 |
| EmilienM | my exec is really straightforward, I think we should move on with it now | 16:05 |
| EmilienM | moreover, I manage the vswitch with puppet | 16:05 |
| EmilienM | in fact, there is no bug (cc _ody), puppet run is idempotent for vswitch module too | 16:05 |
| EmilienM | I think we can move to 'proposed', results are really good: https://review.openstack.org/#/q/status:open++branch:master+topic:beaker/trusty-proposed,n,z | 16:09 |
| sbadia | hi here! | 16:12 |
| * EmilienM afk lunch | 16:17 | |
| _ody | Hunner: I am in the room...when you get onsite. | 16:31 |
| Hunner | nibalizer: Coming in today, or just hanging out? | 16:55 |
| * EmilienM back | 17:28 | |
| EmilienM | pabelanger: can you -1/+1 https://review.openstack.org/#/c/220527/ when you got time? | 17:32 |
| *** mrmartin has quit IRC | 17:37 | |
| EmilienM | we are going to loose network for 15 min - office outage | 17:44 |
| Hunner | Office outage in the middle of the workday? That's odd... | 17:45 |
| Hunner | And sounds expensive | 17:45 |
| EmilienM | crinkle: neutron fixed a bug upstream, that means we can enable again LBaaS https://review.openstack.org/#/c/220566/ | 17:45 |
| crinkle | yay \o/ | 17:46 |
| EmilienM | Hunner: yes... one of the reason I love working from home | 17:46 |
| EmilienM | back online | 18:02 |
| EmilienM | crinkle: in the next cycle, we could even imagine puppet jobs gating (experimental) some other projects, it would be interested | 18:02 |
| EmilienM | interesting* | 18:02 |
| EmilienM | crinkle: the tempest patch is passing on trusty now!! except a bug in tempest that I'm investigating with matthew (tempest ptl) - that means we have integration jobs working on both distros soon :-) | 18:11 |
| nibalizer | Hunner: hi | 18:19 |
| Hunner | nibalizer: hi | 18:19 |
| nibalizer | you at puppetlabs? | 18:19 |
| nibalizer | I'll come down if I'm welcome :) | 18:19 |
| Hunner | Yeah, for a bit. Probably leaving around 1ish | 18:19 |
| Hunner | And lunch at 12ish, so not much time ¯\_(ツ)_/¯ | 18:20 |
| Hunner | But you're welcome :) | 18:20 |
| Hunner | And _ody is here | 18:20 |
| *** mrmartin has joined #openstack-sprint | 18:23 | |
| nibalizer | Hunner: aight i'll bomb on down | 18:26 |
| pabelanger | http://logs.openstack.org/97/220297/15/experimental/gate-puppet-openstack-integration-dsvm-multinode-nv/af03af6/console.html | 18:30 |
| pabelanger | using ansible to invoke zuul-cloner / r10k on the remote node | 18:30 |
| pabelanger | works like expected | 18:30 |
| pabelanger | EmilienM: ^ | 18:30 |
| EmilienM | excellent :) | 18:31 |
| EmilienM | crinkle: can you review https://review.openstack.org/#/c/219818/ please ? | 19:10 |
| crinkle | EmilienM: I thought openstack_extras::repo took care of installing the keyring and apt-get updating? if not it's not working properly | 19:11 |
| EmilienM | crinkle: openstack_extras::repo runs apt-get update *after* adding the keyring package | 19:12 |
| EmilienM | crinkle: we need to do that in integ because trying to install a package without apt-get update provides a race condition | 19:12 |
| EmilienM | sometimes it works | 19:12 |
| EmilienM | sometimes not | 19:12 |
| EmilienM | and we don't have the issue in beaker because beaker runs apt-get update for us | 19:12 |
| EmilienM | when preparing the node | 19:12 |
| crinkle | hrm okay | 19:13 |
| crinkle | EmilienM: what's the deal with ceilometer and horizon on centos? | 19:14 |
| EmilienM | crinkle: I see no clean way to tell puppet running it before, then after again.. except adding an exec somewhere | 19:14 |
| EmilienM | crinkle: https://etherpad.openstack.org/p/puppet-liberty-blocker | 19:14 |
| EmilienM | https://bugs.launchpad.net/puppet-ceilometer/+bug/1490986 | 19:14 |
| openstack | Launchpad bug 1490986 in puppet-ceilometer "centos: puppet run is not idempotent (collector & notif agents)" [Critical,In progress] - Assigned to Emilien Macchi (emilienm) | 19:14 |
| EmilienM | horizon is ? RDO horizon folks are on it right now | 19:14 |
| EmilienM | I'm asking to mrunge | 19:15 |
| EmilienM | <mrunge> 3 missing packages | 19:17 |
| EmilienM | <mrunge> jpena already packaged 2 of them | 19:17 |
| EmilienM | crinkle: ^ | 19:17 |
| EmilienM | that means horizon should be fixed very soon | 19:17 |
| EmilienM | and since it's feature freeze, I guess we can expect RDO packages more stable | 19:18 |
| _ody | nibalizer: I am on the 6th floor at my desk. Was just thinking about getting food. Hunner is having lunch with someone today. | 19:21 |
| * _ody didn't see you | 19:25 | |
| EmilienM | I need to jump in my train, probably back online later | 19:26 |
| *** cdelatte has quit IRC | 19:49 | |
| EmilienM | back | 19:56 |
| richm | EmilienM: I don't know if you saw my latest ramblings on #puppet-openstack | 19:57 |
| richm | EmilienM: I've got a big problem with puppet introspection | 19:57 |
| EmilienM | nibalizer, crinkle: do you know why https://review.openstack.org/#/c/219804/ and others do not merge? the patch in Depends-On is merged though, should I trigger something? 'recheck' ? | 19:58 |
| EmilienM | richm: looking now | 19:58 |
| richm | Puppet::Resource.indirection.find is apparently not working if the resource name contains "::"? | 19:58 |
| EmilienM | wow, I never tested it, maybe _ody and Hunner did | 19:58 |
| EmilienM | richm: testing | 19:59 |
| mfisch | sounds like a bug | 19:59 |
| EmilienM | I hope not | 19:59 |
| nibalizer | EmilienM: Ill look | 20:11 |
| nibalizer | EmilienM: it is running tests in gate | 20:12 |
| nibalizer | you can see it at http://status.openstack.org/zuul/ | 20:12 |
| *** asselin_ has joined #openstack-sprint | 20:14 | |
| *** asselin has quit IRC | 20:16 | |
| *** EmilienM has quit IRC | 20:16 | |
| *** EmilienM_ has joined #openstack-sprint | 20:16 | |
| *** EmilienM_ is now known as EmilienM | 20:16 | |
| EmilienM | nibalizer: I +A to trigger gate | 20:17 |
| nibalizer | EmilienM: yes and its running | 20:19 |
| EmilienM | nibalizer: but I was wondering if I had to do that or if I missed something | 20:20 |
| nibalizer | I can't say what state it is in before the second +a was added | 20:21 |
| nibalizer | but I think it doesn't matter | 20:22 |
| nibalizer | EmilienM: pabelanger https://review.openstack.org/#/c/219879/ would add nonvoting puppet4 beaker tests | 20:44 |
| EmilienM | nibalizer: so you want to test 3.x also? | 20:48 |
| *** mrmartin has quit IRC | 20:48 | |
| nibalizer | EmilienM: yes | 20:48 |
| nibalizer | I need to for infra for the forseable future | 20:48 |
| nibalizer | pabelanger: if you're around want to chat about https://review.openstack.org/#/c/220527/4/jenkins/jobs/puppet-module-jobs.yaml ? | 20:49 |
| nibalizer | I think branch inspection inside a job is the path to madness | 20:49 |
| nibalizer | We did it a TON internally at HP and it ended up being super confusing | 20:49 |
| nibalizer | although it is worth noting that infra doesn't write or run unit tests so its all good | 20:50 |
| nibalizer | from infras perspective | 20:50 |
| EmilienM | nibalizer: reviewed | 20:51 |
| EmilienM | nibalizer: I actually prefer your naming, but in that case we should also rename unit & syntax jobs to be consistent | 20:51 |
| EmilienM | nibalizer: have you seen https://review.openstack.org/220527 ? | 20:52 |
| EmilienM | oh sorry | 20:52 |
| EmilienM | just got the buffer (in the train atm) | 20:53 |
| nibalizer | haha | 20:53 |
| nibalizer | are you headed home already? | 20:53 |
| nibalizer | safe travels if so | 20:53 |
| EmilienM | no I have 3 hours of train | 20:53 |
| EmilienM | thanks :) | 20:53 |
| _ody | 3 hours? WHY! | 20:58 |
| *** rfolco has quit IRC | 21:02 | |
| EmilienM | _ody: montreal -> quebec | 21:03 |
| nibalizer | EmilienM: sounds like a good reason to get gertty up | 21:04 |
| EmilienM | yeah :) | 21:04 |
| *** HenryG has quit IRC | 21:07 | |
| *** zaro has quit IRC | 21:09 | |
| EmilienM | Hunner: WOOT !!!!! https://forge.puppetlabs.com/openstack | 21:10 |
| *** HenryG has joined #openstack-sprint | 21:11 | |
| nibalizer | mmm approvals | 21:11 |
| EmilienM | our pupper modules getting approved, awesome | 21:11 |
| *** zaro has joined #openstack-sprint | 21:22 | |
| pabelanger | nibalizer: taking it easy now, but don't mind adding it for experimental for now. If people don't want it there, then Rakefile works, however I don't want to update each module manually, so we're talking about some sort of gem to hold our unit-test layouts. | 21:38 |
| pabelanger | so, can go either way | 21:38 |
| pabelanger | but would rather deal with JJB over a gem presonally | 21:38 |
| pabelanger | longer term, getting -infra involved and doing some sort of single node, that launches chroots / containers for this purpose | 21:39 |
| nibalizer | ya I definitely believe there is a need in -infra for a lightweight node | 21:44 |
| nibalizer | maybe it is a bunch of containers that nodepool can give out | 21:44 |
| nibalizer | because things like lint and syntax don't require the huge nodes that infra usually deals with | 21:45 |
| pabelanger | exactly | 21:45 |
| nibalizer | at which point having separate jobs is the way to go | 21:46 |
| EmilienM | nibalizer: we made it to save nodepool resources | 21:47 |
| EmilienM | probably we can improve it | 21:47 |
| clarkb | nibalizer: is there a concern with using nodes that are too big? | 21:48 |
| * clarkb is somewhat confused but in general the way we handle this elsewhere is run tests in parallel per run | 21:49 | |
| clarkb | uses all the avialable resources as quickly as possible | 21:49 |
| nibalizer | clarkb: so at this point we are using 16 nodes per puppet-* change | 21:53 |
| nibalizer | I haev a change in flight to increase that by 6 | 21:53 |
| nibalizer | so we're trying to bring that number down | 21:53 |
| nibalizer | the goal is to reduce the impact to infra | 21:53 |
| clarkb | I thought it was like 6? lint + all the version unittests + a rspec breaker run or two | 21:53 |
| nibalizer | https://review.openstack.org/#/c/219289/ | 21:53 |
| clarkb | nibalizer: so first thing I would do is smash lint and syntax together since they are all common static analysis | 21:54 |
| nibalizer | and I have a change up to run beaker on puppet3 and puppet 4 | 21:54 |
| nibalizer | clarkb: ya paul has a change up for that | 21:54 |
| nibalizer | thats actually what we are discussing | 21:54 |
| nibalizer | https://review.openstack.org/220527 | 21:54 |
| clarkb | and I would drop the non upgrade case becaues thats a subset of the upgrade case | 21:54 |
| nibalizer | it gets a little complicated | 21:54 |
| nibalizer | oh hrm | 21:55 |
| nibalizer | I'm not sure if that is true but it might be | 21:55 |
| clarkb | and finally I would pick puppet versions that exist on distros or other common deployment and not do all of them | 21:55 |
| clarkb | kind of like how we do python | 21:55 |
| nibalizer | ya I think doing 3.4 3.x and 4.x is probably sufficient | 21:55 |
| clarkb | I don't know what those versions are but testing every version seems silly now particularly that puppetlabs says we are sember | 21:55 |
| clarkb | *semver | 21:55 |
| clarkb | if that breaks file a bug with puppetlabs move on | 21:56 |
| nibalizer | well if someone starts using a feature that was added in 3.5 we need to test on 3.4 to know that it breaks 3.4 compat | 21:56 |
| clarkb | that gets you to like 6 jobs | 21:56 |
| clarkb | nibalizer: then test the earliest 3 version and the earliest 4 version? | 21:56 |
| clarkb | anyways, my point is that rather than getting smart about how we run jobs and complicating the tooling we should instead get smart about what is worth testing and how we test then go from there | 21:57 |
| nibalizer | sure that works today | 21:57 |
| EmilienM | nibalizer: yes | 21:57 |
| EmilienM | it would be imho | 21:57 |
| nibalizer | but why not have infra provide lightweight nodes so that we can run lots of cheap tests without the complexity of chaining them | 21:58 |
| clarkb | the test platform is somewhat intentionally dumb and we are trying to make it even simpler | 21:58 |
| nibalizer | im not say we do that today | 21:58 |
| nibalizer | but in the future | 21:58 |
| clarkb | nibalizer: because that makes the system complicated. Not saying we can't do it if there is a real need but I like simple | 21:58 |
| clarkb | and right now I honestly don't think that puppet modules are any more special than everything else we test | 21:59 |
| clarkb | which we have done effectively with the nodes available | 21:59 |
| nibalizer | I think this problem exists across the board | 21:59 |
| nibalizer | where some tests need big nodes and some do not | 21:59 |
| clarkb | but thats ok | 21:59 |
| clarkb | there is never a case where running a small test on a big node is problematic | 21:59 |
| clarkb | if you need to restrict resources to keep a test case small that is doable without extra special node types | 22:00 |
| nibalizer | sure but if the quota infra has from cloud providers is the limiting factor, more smaller nodes could exist than big nodes right? | 22:00 |
| nibalizer | or is the api rate limiting the slow part? | 22:00 |
| clarkb | nibalizer: we can query graphite, its both to an extent | 22:01 |
| clarkb | nibalizer: when we aren't at quota we really feel the rate limiting (whther by api servers or naturally due to delays in operations) | 22:01 |
| nibalizer | ya | 22:01 |
| clarkb | when we are at quota we of course can't boot more things | 22:01 |
| clarkb | but for example in hpcloud we can't boot more nodes even if we made them smaller so thats not a huge win | 22:02 |
| nibalizer | it looks like we got near quota today | 22:02 |
| nibalizer | but that has stopped | 22:02 |
| clarkb | in rax thats the same too, would require much more tuning in quota | 22:02 |
| clarkb | so you would hvae the same number of nodes they would just be less useful | 22:02 |
| clarkb | and your shcedulre would have to be smarter to place jobs on the appropriate node sizes | 22:02 |
| nibalizer | so I think having more small nodes is only valuable when we're near quota | 22:02 |
| clarkb | and to know when to boot which type properly | 22:02 |
| nibalizer | clarkb: you can do that with label right? | 22:03 |
| clarkb | nibalizer: except for the fact that we have quota limits other than ram and cpu | 22:03 |
| nibalizer | label is trusty-small or trusty-dsvm | 22:03 |
| clarkb | nibalizer: sort of, I mean nodepool does it but it isn't always great at it | 22:03 |
| nibalizer | so a lint check goes on a small | 22:03 |
| clarkb | nibalizer: right the hard part is when you have 100 small jobs, 100 big jobs, and 25 big quota/40small quota | 22:03 |
| nibalizer | and of course, not to be this guy, but four big nodes in rax that just host docker containers wouldn't be affected by cloud api lmits | 22:03 |
| clarkb | how do you fit into those buckets? | 22:03 |
| clarkb | pretty sure solving these problems is what people do toget PhDs | 22:04 |
| nibalizer | clarkb: ya thats a hard problem | 22:04 |
| nibalizer | beyond me | 22:04 |
| nibalizer | but right now we put a 30G node in charge of running puppet-lint | 22:04 |
| nibalizer | which seems silly | 22:04 |
| clarkb | so from my perspetive supporting the simple thing is easier for everyone | 22:04 |
| clarkb | nibalizer: its an 8GB 8vcpu node | 22:04 |
| nibalizer | i thought it was 30G then gimped to 8 | 22:05 |
| clarkb | nibalizer: what the underlying cloud flavor is is an implementation detail, the tests are unaware | 22:05 |
| clarkb | I suppose they could discover that the kernel boot parameters restrict memory, update the kernel boot param, reboot then the test would fail | 22:08 |
| clarkb | but that would be the only case any specific test is aware of such things | 22:08 |
| pabelanger | Ya simple is great, and I love how our -infra works today. But I agree with nibalizer on that point, launching multiple nodes for lint checks is over kill in that instance. I think something could be added, but not sure what that looks like. What I did today, was a simple first iteration | 22:13 |
| pabelanger | I also agree with clarkb that we could prune the puppet versions to what distros ship today | 22:14 |
| pabelanger | train time | 22:15 |
| pabelanger | will be back in -infra | 22:16 |
| *** pabelanger has left #openstack-sprint | 22:16 | |
| clarkb | but ya as a today note our quota allows the number of hosts we run today. It would not allow more nodes just because we made msaller nodes | 22:19 |
| clarkb | so there is literally no benefit to mixing smaller nodes in | 22:20 |
| crinkle | is there a problem for us on the puppet side that reducing the number of nodes fixes? | 22:22 |
| crinkle | seems like we're trying to help Infra out and Infra is saying they don't need the help | 22:23 |
| clarkb | well I am saying that using smaller nodes won't help. Using fewer nodes may help particularly when we are at quota | 22:23 |
| crinkle | okay | 22:23 |
| clarkb | my suggestions were a possible way of doing that without losing test coverage, but I don't know enough about your testing to be sure | 22:24 |
| nibalizer | clarkb: see the patch the paul made | 22:24 |
| nibalizer | we beed different thibgs to happen based on branches | 22:24 |
| nibalizer | so ita complex | 22:24 |
| nibalizer | clarkb: im not saying i need smaller nodes, im asking why it wont help and i thibk youve explained that to me | 22:25 |
| nibalizer | maybe we run the normal puppet tests in one test and the 'latest' on a separate test | 22:26 |
| nibalizer | that way we get less nodes used | 22:26 |
| nibalizer | but also have less complexity | 22:26 |
| EmilienM | mfisch: I got issues to spawn a VM on ubuntu, http://logs.openstack.org/52/217352/56/check/gate-puppet-openstack-integration-dsvm-trusty/b44e53b/logs/nova/nova-compute.txt.gz#_2015-09-04_21_50_18_643 | 22:31 |
| EmilienM | have you seen that before? | 22:31 |
| EmilienM | I might need to adjust something | 22:31 |
| EmilienM | I think dbus needs to be managed | 22:42 |
| EmilienM | that's a bug in puppet-nova | 22:46 |
| EmilienM | thanks everyone for joining the sprint, mfisch, clarkb, crinkle, clayton, nibalizer, Vinsh, richm, _ody, mwhahaha, Hunner, anteaya - I'll make sure we do a retrospective to see how we can do better next time | 23:03 |
| EmilienM | have a great week end | 23:03 |
| *** greghaynes has joined #openstack-sprint | 23:30 | |
| nibalizer | woooo | 23:50 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!