| *** michael-beaver has quit IRC | 00:12 | |
| *** tributarian has quit IRC | 00:33 | |
| *** ianychoi has joined #zuul | 00:34 | |
| *** rlandy is now known as rlandy|bbl | 00:35 | |
| *** bhavikdbavishi has joined #zuul | 01:54 | |
| *** bhavikdbavishi1 has joined #zuul | 01:57 | |
| *** jamesmcarthur has joined #zuul | 01:58 | |
| *** jamesmcarthur has quit IRC | 01:58 | |
| *** jamesmcarthur has joined #zuul | 01:58 | |
| *** bhavikdbavishi has quit IRC | 01:58 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 01:58 | |
| *** rlandy|bbl is now known as rlandy | 02:07 | |
| *** rlandy has quit IRC | 02:29 | |
| *** sileht has quit IRC | 02:44 | |
| *** bhavikdbavishi has quit IRC | 02:48 | |
| *** bhavikdbavishi has joined #zuul | 03:33 | |
| *** jamesmcarthur has quit IRC | 03:46 | |
| *** jamesmcarthur has joined #zuul | 03:46 | |
| *** jamesmcarthur has quit IRC | 03:51 | |
| *** jamesmcarthur has joined #zuul | 04:16 | |
| *** jank has joined #zuul | 04:37 | |
| openstackgerrit | Jan Kubovy proposed zuul/zuul master: Make tenant and pipeline optional in zuul-changes https://review.opendev.org/674034 | 04:50 |
|---|---|---|
| *** jamesmcarthur has quit IRC | 04:53 | |
| *** bjackman has joined #zuul | 05:44 | |
| *** jhesketh has joined #zuul | 06:54 | |
| *** jpenag is now known as jpena | 07:31 | |
| openstackgerrit | Merged zuul/nodepool master: builder: Log all deletions of image upload records https://review.opendev.org/674126 | 07:32 |
| *** fbo has joined #zuul | 07:48 | |
| *** tosky has joined #zuul | 07:48 | |
| *** mhu has joined #zuul | 08:10 | |
| *** threestrands has quit IRC | 08:25 | |
| cixx_ | "Zuul is a proxy and therefore benchmarks -- especially RPS-based ones -- could vary widely." what does RPS stand for? | 08:39 |
| aspiers | mordred: did this eatmydata research ever get done? http://eavesdrop.openstack.org/irclogs/%23zuul/%23zuul.2017-03-21.log.html#t2017-03-21T11:28:24 | 08:50 |
| AJaeger | cixx_: that is a quote from netflix Zuul which is a complete different tool, they "stole" our name ;) Zuul is https://zuul-ci.org/ | 09:00 |
| *** bhavikdbavishi has quit IRC | 09:13 | |
| *** hwangbo has quit IRC | 09:20 | |
| *** jangutter has quit IRC | 09:36 | |
| *** electrofelix has joined #zuul | 09:37 | |
| *** jangutter has joined #zuul | 09:51 | |
| *** jank has quit IRC | 10:00 | |
| *** jank has joined #zuul | 10:07 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: [WIP] Add OpenAPI description for enqueue, dequeue, autohold https://review.opendev.org/674257 | 10:25 |
| *** jank has quit IRC | 10:27 | |
| *** zbr has quit IRC | 11:25 | |
| *** jpena is now known as jpena|lunch | 11:34 | |
| *** SotK__ has quit IRC | 11:51 | |
| *** tributarian has joined #zuul | 11:59 | |
| *** panda is now known as panda|lunch | 12:05 | |
| *** rlandy has joined #zuul | 12:25 | |
| *** zbr has joined #zuul | 12:35 | |
| *** jpena|lunch is now known as jpena | 12:39 | |
| *** jamesmcarthur has joined #zuul | 12:39 | |
| cixx_ | AJaeger: I know :) I just wanted to try my chance here to find what RPS stands for | 12:44 |
| jangutter | cixx_: Rock Paper Scissors? | 12:49 |
| *** bjackman has quit IRC | 12:55 | |
| *** panda|lunch is now known as panda | 13:06 | |
| *** jeliu_ has joined #zuul | 13:12 | |
| fungi | cixx_: reverse proxy server | 14:01 |
| fungi | as opposed to a normal (forward) proxy | 14:01 |
| fungi | if you stick a caching proxy cdn in front of a site, for example, that's an "rps" | 14:02 |
| fungi | because it's a proxy tied to the server side of the connection, not the client side | 14:02 |
| fungi | so masquerading as the server itself in most cases | 14:02 |
| *** jeliu_ has quit IRC | 14:04 | |
| *** jamesmcarthur has quit IRC | 14:10 | |
| *** jeliu_ has joined #zuul | 14:11 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: Add OpenAPI description for enqueue, dequeue, autohold https://review.opendev.org/674257 | 14:23 |
| *** jamesmcarthur has joined #zuul | 14:24 | |
| *** electrofelix has quit IRC | 14:25 | |
| *** jamesmcarthur has quit IRC | 14:30 | |
| tobiash | aspiers: we're using it in quite some jobs to save iops | 14:30 |
| aspiers | tobiash: nice! | 14:30 |
| tobiash | depending on how much ram is available this can make quite some difference (depending on the job) | 14:31 |
| tobiash | aspiers: how did you find that old cite? ;) | 14:31 |
| aspiers | 14:32 | |
| *** bmwiedemann1 has joined #zuul | 14:33 | |
| *** jamesmcarthur has joined #zuul | 14:33 | |
| jangutter | fungi: The archaic RPS I know is for "receive packet steering" a.k.a the thing you use if you don't have hardware multiqueue NICs. | 14:37 |
| *** michael-beaver has joined #zuul | 14:47 | |
| *** bmwiedemann1 has left #zuul | 14:56 | |
| *** jeliu_ has quit IRC | 15:02 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: WIP: add-build-sshkey: add centos/rhel-8 support https://review.opendev.org/674092 | 15:03 |
| *** jeliu_ has joined #zuul | 15:04 | |
| clarkb | zbr: I left a question and a thought on ^ | 15:20 |
| clarkb | aspiers: fwiw in openstack land I've been poking at job slowness and posting results to the openstack-discuss mailing list. One of the issues is jobs are swapping | 15:21 |
| clarkb | aspiers: I'm not sure eatmydata will help much in those cases | 15:21 |
| aspiers | k | 15:21 |
| clarkb | aspiers: http://lists.openstack.org/pipermail/openstack-discuss/2019-July/008071.html and http://lists.openstack.org/pipermail/openstack-discuss/2019-July/007493.html for more info | 15:22 |
| openstackgerrit | Merged zuul/zuul-jobs master: Don't compare to literal True/False https://review.opendev.org/667697 | 15:22 |
| openstackgerrit | Merged zuul/zuul-jobs master: Add test-bindep job https://review.opendev.org/674078 | 15:22 |
| *** mattw4 has joined #zuul | 15:22 | |
| openstackgerrit | Merged zuul/nodepool master: builder: Remove recency table logging https://review.opendev.org/674124 | 15:30 |
| *** sileht has joined #zuul | 15:34 | |
| *** mattw4 has quit IRC | 15:38 | |
| zbr | clarkb: thanks for the comment. re ansible boostrapping, its is much trickier than you think. | 15:41 |
| zbr | clarkb: is about bootstrapping what, there are lots of requirements, most of them loose. | 15:41 |
| zbr | clarkb: also a well written role should list *all* requirements in its meta (dependencies). currently we do not make use of that, but this does not mean we should not aim to. | 15:42 |
| clarkb | zbr: well for example we must have a python installed on the remote node | 15:42 |
| clarkb | opendev ensures that happens at image build time | 15:42 |
| Shrews | ianw: fyi, i went ahead and abandoned https://review.opendev.org/630890 for you as we no longer use the devstack plugin | 15:43 |
| clarkb | part of my concern here is the chicken and egg problem | 15:43 |
| clarkb | I don't think we can assume ansible can bootstrap its own deps (for example python) | 15:43 |
| clarkb | and so some external process is already doing that may as well deal with them all there? | 15:43 |
| zbr | clarkb: that is something opaque and not tested, i doubt all these roles are tested on each image build. | 15:43 |
| clarkb | well right now we have no rhel8 or centos 8 image | 15:44 |
| clarkb | you are correct that isn't tested | 15:44 |
| clarkb | I'm talking about this in the general case | 15:44 |
| zbr | clarkb: you can use ansible to bootstrap ansible on a machine, but is not an easy task. (you may need to use raw command before you get to python) | 15:44 |
| clarkb | if every role that uses authorized_key module must also include tasks to intsall selinux python bindings I think we have done something wrong | 15:44 |
| zbr | clarkb: i am trying to improve testing and slowly making these roles easy to re-use in various places. | 15:45 |
| *** jpena is now known as jpena|off | 15:45 | |
| *** mhu has quit IRC | 15:45 | |
| openstackgerrit | Luigi Toscano proposed zuul/zuul-jobs master: fetch-subunit-output: collect additional subunits (2nd try) https://review.opendev.org/674334 | 15:46 |
| tosky | let's retry! | 15:46 |
| *** mhu has joined #zuul | 15:46 | |
| clarkb | zbr: is the typical ansible role expected to bootstrap all of its dependencies like that? | 15:47 |
| zbr | clarkb: and with some guidance we may succeed. I am still learning about how test-zuul.d is supposed to be used. I am used to a very different approach: using molecule to test roles and testing them in containers. this means that I could run the tests locally with tox without needing zuul. | 15:47 |
| AJaeger | tosky: sure ;) Could you add a requirements or openstacksdk change and make it use the updated job, please? | 15:47 |
| pabelanger | clarkb: zbr: history has taught me, to manage ansible dependencies in a single location, rather then having each role trying individual add things they need. | 15:47 |
| clarkb | maybe that is what I am asking. What is the expected process here. Because to me installing every minor ansible dep in every role that needs them seems like a lot of unnecessary effort | 15:47 |
| tosky | AJaeger: any of them? | 15:47 |
| zbr | clarkb: yes/not: mainly you only need to list bootstraping roles as dependencies. ansible will run them only once by default. | 15:47 |
| AJaeger | tosky: either one should be enough - or choice ;) | 15:47 |
| tosky | oki | 15:48 |
| zbr | clarkb: I could adda a 'bootstrap' role to the repository that does exactly these things, and we can start referring to it from the other ones. | 15:48 |
| clarkb | zbr: I think that would be preferable | 15:48 |
| zbr | mainly the boostraping role would be very similar with what you do to build the images. | 15:49 |
| clarkb | otherwise we will have a proliferation of these python lib installs | 15:49 |
| zbr | clarkb: any preference for the role name? | 15:49 |
| clarkb | install-ansible-remote-deps ? something like that | 15:49 |
| zbr | clarkb: can I use molecule to test it? we can test it on 5 distros with a single job (far smaller resource usage). | 15:50 |
| clarkb | I have no opinon on whether or not molecule should be used | 15:51 |
| *** jangutter has quit IRC | 15:51 | |
| zbr | sure, i can do both and others can decide. also adding molecule files can help a lot when making changes locally, regardless if is used by zuul or not. | 15:52 |
| zbr | i would prefer to name the role bootstrap-ansible, is shorter and this is what it would do: asure that the host can be managed by ansible (basic modules). In the future we can add extras. | 15:54 |
| zbr | but libselinux is something we need almost anywhere, you can barely run ansible without them. | 15:54 |
| zbr | jinja templating requires it | 15:54 |
| *** mattw4 has joined #zuul | 15:57 | |
| *** hwangbo has joined #zuul | 16:05 | |
| *** jamesmcarthur has quit IRC | 16:08 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: Add CORS support to upload-logs-swift https://review.opendev.org/674341 | 16:08 |
| tosky | AJaeger: https://review.opendev.org/#/c/674344/ | 16:12 |
| *** pcaruana has quit IRC | 16:14 | |
| openstackgerrit | Merged zuul/zuul-jobs master: Support Rackspace in upload-logs-swift https://review.opendev.org/674136 | 16:18 |
| *** altlogbot_3 has quit IRC | 16:29 | |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Be consistent about spaces before and after vars https://review.opendev.org/667698 | 16:31 |
| *** irclogbot_0 has quit IRC | 16:33 | |
| AJaeger | tosky: thanks, please add as comment to the other change | 16:35 |
| *** altlogbot_0 has joined #zuul | 16:38 | |
| tosky | and done | 16:38 |
| *** tributarian has quit IRC | 16:38 | |
| *** mattw4 has quit IRC | 16:42 | |
| *** mattw4 has joined #zuul | 16:43 | |
| *** irclogbot_1 has joined #zuul | 16:43 | |
| corvus | zbr: zuul-jobs has facilities for testing on multiple repos; i don't think we should add molecule into the mix | 16:43 |
| corvus | (we *just* finished setting up the testing framework for zuul-jobs, i think it's maybe a bit early to replace it) | 16:44 |
| corvus | zbr: see https://zuul-ci.org/docs/zuul-jobs/policy.html#testing for how to do multi-platform testing | 16:44 |
| corvus | clarkb: ^ fyi | 16:44 |
| corvus | oh, this is for add-build-sshkey.... | 16:47 |
| corvus | zbr, clarkb: testing that role is going to be very difficult, if it's possible at all. it has some pretty low-level interaction with zuul | 16:49 |
| *** mattw4 has quit IRC | 16:50 | |
| corvus | it is responsible for dropping the global ssh key; even in base-minimal we do that because it's a security vulnerability not to | 16:50 |
| clarkb | my concern was that we not start adding ansible dependency bootstrapping to every role that might interact iwth selinux on selinux platforms | 16:50 |
| clarkb | which is orthogonal to testing | 16:50 |
| corvus | clarkb: i know, i read that | 16:50 |
| clarkb | as far as testing that particular role goes ya I can see how changing ssh keys would make zuul/ansible potentially unhappy | 16:51 |
| clarkb | I wonder how valid the test is if we replace the build key with the build key | 16:51 |
| corvus | but zbr asked about molecule, and you didn't have anything to add there, so i thought i should add information now, rather than after zbr went and did a bunch of stuff that we might not accept | 16:51 |
| clarkb | ++ | 16:53 |
| corvus | the biggest hurdle i see to testing it is this task: https://opendev.org/zuul/zuul-jobs/src/branch/master/roles/add-build-sshkey/tasks/create-key-and-replace.yaml#L39-L45 | 16:53 |
| clarkb | (I've never used molecule and know little about it so hard to offer opinions on it) | 16:53 |
| corvus | that task responds to a state created by the zuul-executor which i think is impossible to recreate or simulate in a direct test | 16:54 |
| clarkb | ah | 16:54 |
| corvus | i think the only way to test it would be in a recursive situation where we run a second ansible (with its own ssh agent) | 16:54 |
| corvus | sort of like how we do the console output tests on zuul itself | 16:55 |
| corvus | i think molecule is great -- the reason we don't use it in zuul-jobs is that we're mostly trying to test interaction with zuul; so test isolation is an anti-goal | 16:55 |
| corvus | but if it is appropriate for a particular role (maybe it would help in this case) i think we should consider it. just that on balance, most of the roles are better tested in situ. | 16:57 |
| openstackgerrit | Merged zuul/zuul-jobs master: Add CORS support to upload-logs-swift https://review.opendev.org/674341 | 16:59 |
| fungi | i can see where molecule might be useful for testing in-repo roles which projects want to serve double-duty in both testing and deployment | 16:59 |
| openstackgerrit | Sorin Sbarnea proposed zuul/zuul-jobs master: Make all lines less than 160 characters long https://review.opendev.org/667696 | 17:00 |
| fungi | for roles which are only intended to be run as part of zuul jobs, it's probably not a good fit | 17:00 |
| clarkb | I do wish these config management testing tools supported "run tests against the local system" as an expected workflow | 17:00 |
| clarkb | molecule isn't the first one to ignore this use case | 17:00 |
| *** igordc has joined #zuul | 17:03 | |
| fungi | for that matter, you could say tox has that issue too | 17:05 |
| fungi | (though maybe there's a tox option to not create a virtualenv) | 17:05 |
| clarkb | fungi: https://pypi.org/project/tox-direct/ you can do that via a plugin | 17:08 |
| clarkb | (so ya generally suffers the same problem) | 17:09 |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 17:09 |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 17:12 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: DNM: test swift logs https://review.opendev.org/674358 | 17:15 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: DNM: test swift logs https://review.opendev.org/674359 | 17:16 |
| *** panda has quit IRC | 17:19 | |
| fungi | ahh, neat, i was unaware of tox-direct | 17:19 |
| fungi | though on the topic of tox plugins, i rather enjoy tox-venv in my python3-only projects | 17:19 |
| fungi | frees you from needing to use virtualenv, which is becoming increasingly problematic these days | 17:20 |
| *** panda has joined #zuul | 17:20 | |
| *** openstackgerrit has quit IRC | 17:22 | |
| zbr | clarkb: fyi: if you want to run molecule on localhost: you just need to use "delegated" driver. which mainly allows you to give it an inventory to use. | 17:22 |
| zbr | if you put localhost on it... it will run on localhost. we already do this in few jobs. still, as one could imagine, this has the potential to f*** you localhost. | 17:23 |
| clarkb | zbr: ya but it is typically what you want in a ci environment | 17:23 |
| clarkb | because the environment is provisioned for you | 17:23 |
| zbr | clarkb: yep. but when I think about functional testing, I try to find a way that works for both local-dev-env and CI. | 17:24 |
| *** mattw4 has joined #zuul | 17:24 | |
| zbr | so if I can isolate it, the better. developer can test locally, in seconds and avoid proposing broken patches. | 17:24 |
| zbr | and as a note: one way of doing it does not rule out the other. | 17:25 |
| clarkb | zbr: that works great if you can get away with containers but if you need virtual machines nesting them is bad | 17:26 |
| zbr | clarkb: true. but so far i was able to get around with containers for big number of cases. | 17:27 |
| clarkb | so it is mostly about having the option not necessarily that you must use it | 17:27 |
| zbr | yep | 17:28 |
| zbr | clarkb: i am starting to like the guy who wrong tox-direct, he has humor: --direct-yolo | 17:33 |
| zbr | yolo mode :D | 17:33 |
| fungi | heh | 17:35 |
| zbr | btw, did anyone tried to use container based nodesets with zuul? from what I read it it should be supported. for tox jobs they could present some benefits. | 17:35 |
| clarkb | zbr: I believe that softwarefactory may do so. For opendev the way our quotas are given to us some clouds have expressed hesitation over having always on servers providing container resources | 17:37 |
| fungi | i assume the folks using openshift and kubernetes node drivers are using container-based nodesets with zuul? | 17:37 |
| fungi | also the napkin math on overhead we would actually save by switching opendev's nodesets to containers is negligible. focusing on improving the efficiency of the test payloads themselves would be a much better payoff for the time spent | 17:38 |
| fungi | (efficiency gains in tripleo-ci and devstack/tempest being the obvious low-hanging fruit there) | 17:39 |
| corvus | still, i'd love to have some testing of it in opendev, but it doesn't seem like anyone has the time to drive setting that up at the moment | 17:40 |
| fungi | yes, exercising the feature would certainly be nice | 17:40 |
| corvus | https://review.opendev.org/620756 is where that stands | 17:41 |
| corvus | probably need a new k8s cluster at this point | 17:41 |
| *** rfolco|ruck has quit IRC | 17:42 | |
| clarkb | which might be complicated by magnum's recent problems deploying newer k8s (assuming we want to reuse magnum for that) | 17:43 |
| *** openstackgerrit has joined #zuul | 17:47 | |
| openstackgerrit | James E. Blair proposed zuul/zuul master: WIP: render console in js https://review.opendev.org/674368 | 17:47 |
| corvus | mordred, tristanC: ^ that's a start on the long-envisioned "render the console log in JS with collapsable sections" | 17:47 |
| corvus | just a skeleton so far | 17:48 |
| openstackgerrit | James E. Blair proposed zuul/zuul master: WIP: render console in js https://review.opendev.org/674368 | 17:50 |
| *** fdegir has quit IRC | 17:55 | |
| *** fdegir has joined #zuul | 17:56 | |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 18:06 |
| *** cixx_ has quit IRC | 18:21 | |
| *** cixx has joined #zuul | 18:23 | |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 18:35 |
| tristanC | corvus: shouldn't the job-output rendering be part of the summary tab? | 18:46 |
| tristanC | the current summary table could probably takes less space, leaving room for more interesting content | 18:47 |
| *** rfolco has joined #zuul | 19:00 | |
| *** rfolco is now known as rfolco|ruck | 19:01 | |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 19:09 |
| *** portdirect has quit IRC | 19:13 | |
| *** portdirect has joined #zuul | 19:13 | |
| *** tosky has quit IRC | 19:32 | |
| *** mgoddard has quit IRC | 19:37 | |
| *** mgoddard has joined #zuul | 19:38 | |
| *** irclogbot_1 has quit IRC | 19:39 | |
| corvus | tristanC: that ps leave the current output (which tries to find the error -- i really like that) as part of the summary tab, the full console is in a new tab. maybe we could integrate the two? at any rate, i figured this is a good start for experimenting. | 19:40 |
| *** irclogbot_1 has joined #zuul | 19:44 | |
| tristanC | corvus: i also having the potential error first, but it would be nice to dig in plays and task without having to change pane. perhaps the summary table could be moved to another pane, or displayed as some sort of info box on the top right | 19:44 |
| tristanC | also like* | 19:44 |
| corvus | tristanC: that makes sense. once we get the console rendering done, we can explore those options. | 19:50 |
| corvus | there's a bunch of formatting that needs to be done (right now, it mostly just looks like an incomplete list of tasks :) | 19:50 |
| openstackgerrit | Jeff Liu proposed zuul/zuul-operator master: WIP: Add zuul-operator-functional-openshift job https://review.opendev.org/674355 | 19:54 |
| mordred | corvus: ooh! that's exciting | 20:06 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: Fix CORS in rackspace in upload-logs-swift https://review.opendev.org/674423 | 20:39 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: Fix CORS in rackspace in upload-logs-swift https://review.opendev.org/674423 | 20:40 |
| *** jeliu_ has quit IRC | 20:59 | |
| openstackgerrit | Mark Meyer proposed zuul/zuul master: Rework some bugs https://review.opendev.org/674425 | 21:02 |
| *** EmilienM is now known as EmilienM|afk | 21:28 | |
| *** EmilienM|afk is now known as EmilienM | 21:28 | |
| *** rlandy has quit IRC | 21:31 | |
| openstackgerrit | Merged zuul/zuul-jobs master: Fix CORS in rackspace in upload-logs-swift https://review.opendev.org/674423 | 21:41 |
| *** armstrongs has joined #zuul | 22:25 | |
| *** armstrongs has quit IRC | 22:35 | |
| *** hwangbo has quit IRC | 23:26 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!