| *** y2kenny has quit IRC | 00:03 | |
| *** rfolco has quit IRC | 00:29 | |
| openstackgerrit | Clark Boylan proposed zuul/nodepool master: Add basic max-servers handling to the k8s driver https://review.opendev.org/743424 | 00:52 |
|---|---|---|
| *** wuchunyang has joined #zuul | 02:17 | |
| *** sanjayu__ has joined #zuul | 02:22 | |
| *** wuchunyang has quit IRC | 02:35 | |
| *** Goneri has quit IRC | 02:46 | |
| *** y2kenny has joined #zuul | 03:25 | |
| *** bhavikdbavishi has joined #zuul | 03:33 | |
| *** bhavikdbavishi1 has joined #zuul | 03:36 | |
| *** bhavikdbavishi has quit IRC | 03:38 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 03:38 | |
| *** wuchunyang has joined #zuul | 04:00 | |
| *** wuchunyang has quit IRC | 04:04 | |
| *** bhavikdbavishi has quit IRC | 04:22 | |
| *** bhavikdbavishi has joined #zuul | 04:23 | |
| *** evrardjp has quit IRC | 04:33 | |
| *** evrardjp has joined #zuul | 04:33 | |
| *** wuchunyang has joined #zuul | 04:39 | |
| *** y2kenny has quit IRC | 04:50 | |
| *** chandankumar is now known as chkumar|ruck | 05:37 | |
| *** reiterative has quit IRC | 05:41 | |
| *** wuchunyang has quit IRC | 05:41 | |
| *** reiterative has joined #zuul | 05:41 | |
| *** vishalmanchanda has joined #zuul | 05:49 | |
| *** chkumar|ruck is now known as chkumar|rover | 06:13 | |
| *** tosky has joined #zuul | 06:32 | |
| *** zer0c00l has joined #zuul | 06:40 | |
| openstackgerrit | Felix Edel proposed zuul/zuul master: PF4: Fix scrolling behaviour when using keyboard https://review.opendev.org/743239 | 06:44 |
| openstackgerrit | Felix Edel proposed zuul/zuul master: PF4: Use a ref to focus main content on initial page load https://review.opendev.org/743917 | 06:44 |
| *** bhavikdbavishi has quit IRC | 06:46 | |
| *** sgw1 has quit IRC | 06:50 | |
| *** jcapitao has joined #zuul | 07:11 | |
| *** bhavikdbavishi has joined #zuul | 07:16 | |
| *** bhavikdbavishi1 has joined #zuul | 07:20 | |
| *** bhavikdbavishi has quit IRC | 07:21 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 07:21 | |
| *** bhagyashris is now known as bhagyashris|lunc | 07:28 | |
| openstackgerrit | Felix Edel proposed zuul/zuul master: PF4: Use a ref to focus main content on initial page load https://review.opendev.org/743917 | 07:35 |
| openstackgerrit | Guillaume Chauvel proposed zuul/zuul master: Tenant reconfiguration: Get ref-updated modified files https://review.opendev.org/739078 | 07:49 |
| *** jpena|off is now known as jpena | 07:56 | |
| *** sugaar has joined #zuul | 07:56 | |
| *** nils has joined #zuul | 07:56 | |
| *** harrymichal has joined #zuul | 08:29 | |
| *** bhavikdbavishi has quit IRC | 08:30 | |
| *** bhavikdbavishi has joined #zuul | 08:36 | |
| *** harrymichal has quit IRC | 08:48 | |
| *** harrymichal has joined #zuul | 08:48 | |
| *** harrymichal has quit IRC | 08:57 | |
| *** harrymichal has joined #zuul | 08:58 | |
| *** harrymichal has quit IRC | 09:02 | |
| *** harrymichal has joined #zuul | 09:13 | |
| *** hashar has joined #zuul | 09:18 | |
| *** sshnaidm|afk is now known as sshnaidm | 09:20 | |
| *** harrymichal has quit IRC | 09:25 | |
| *** harrymichal has joined #zuul | 09:25 | |
| *** jcapitao is now known as jcapitao_afk | 09:32 | |
| *** bhagyashris|lunc is now known as bhagyashris | 09:37 | |
| *** harrymichal has quit IRC | 09:38 | |
| *** bhavikdbavishi has quit IRC | 09:41 | |
| *** bhavikdbavishi has joined #zuul | 09:43 | |
| *** bhavikdbavishi1 has joined #zuul | 09:52 | |
| *** bhavikdbavishi has quit IRC | 09:52 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 09:52 | |
| *** bhavikdbavishi has quit IRC | 10:03 | |
| *** jcapitao_afk is now known as jcapitao | 10:03 | |
| *** bhavikdbavishi has joined #zuul | 10:03 | |
| *** bhavikdbavishi has quit IRC | 10:08 | |
| *** hashar has quit IRC | 10:41 | |
| *** jcapitao is now known as jcapitao_lunch | 11:05 | |
| zbr | few days ago i proposed a feature for installing ansible roles/collections that would be very useful for cross repo dependency testing | 11:07 |
| zbr | https://github.com/ansible/ansible/issues/70969 | 11:07 |
| zbr | if you could review it and maybe explain the use-case better then me it would be great | 11:08 |
| zbr | basically I want to make it possible to use depends-on, but we need some changes on galaxy client to make it easy | 11:09 |
| zbr | trying to overcome it with zuul roles would not be easy and would also not sort the issue of testing locally | 11:10 |
| tobiash | corvus: I've discovered a flaw in the current implementation of zuul dequeue. It is a blocking call to sched.dequeue which dequeues an item using the main event loop. The problem with this is that when a reconfig is in progress this will block the whole rpclistener (and thus basically all of zuul-web) until this is finished which can take minutes in some cases. | 11:14 |
| tobiash | https://opendev.org/zuul/zuul/src/branch/master/zuul/rpclistener.py#L91 | 11:14 |
| tobiash | do you have an idea how we could improve that? | 11:15 |
| *** sshnaidm_ has joined #zuul | 11:27 | |
| *** sshnaidm has quit IRC | 11:27 | |
| *** jpena is now known as jpena|lunch | 11:29 | |
| *** _erlon_ has joined #zuul | 11:30 | |
| *** sshnaidm_ is now known as sshnaidm | 11:30 | |
| *** rfolco has joined #zuul | 12:02 | |
| *** hashar has joined #zuul | 12:11 | |
| *** jcapitao_lunch is now known as jcapitao | 12:17 | |
| *** jpena|lunch is now known as jpena | 12:32 | |
| *** bhavikdbavishi has joined #zuul | 12:37 | |
| *** bhavikdbavishi1 has joined #zuul | 12:41 | |
| *** bhavikdbavishi has quit IRC | 12:42 | |
| *** bhavikdbavishi1 is now known as bhavikdbavishi | 12:42 | |
| openstackgerrit | Sorin Sbarnea (zbr) proposed zuul/zuul master: Add user preferences dialog https://review.opendev.org/739482 | 12:51 |
| *** sshnaidm is now known as sshnaidm|mtg | 13:08 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Annotate getChangeByUrl logs with event id https://review.opendev.org/743971 | 13:29 |
| *** Goneri has joined #zuul | 13:52 | |
| *** yoctozepto has quit IRC | 13:54 | |
| *** yoctozepto has joined #zuul | 13:55 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Annotate all canMerge calls with event id https://review.opendev.org/743977 | 13:57 |
| corvus | tobiash: we could make the listener more asynchronous, but we'd probably need a pool to deal with that. we could put the jobs (dequeue, enqueue, etc) that need the main loop in one gearman worker, then put the status handler in another gearman worker, that way dequeue callers still wait, but status updates don't. | 14:06 |
| corvus | tobiash: i think the second idea doesn't sound too hard and will solve the immediate problem of keeping the status page lively. which is good, since that will all change with v4. | 14:07 |
| *** vblando has joined #zuul | 14:09 | |
| tobiash | corvus: yeah, I think I prefer the second idea too for now | 14:11 |
| *** vblando has quit IRC | 14:14 | |
| zbr | tobiash: corvus: the prefs is finally ready at https://review.opendev.org/#/c/739482/ | 14:18 |
| mordred | zbr: nice! | 14:21 |
| mordred | zbr: I verify that it behaves as advertised | 14:22 |
| zbr | once merged, i will start migrating the other changes that needed config | 14:22 |
| *** bhagyashris is now known as bhagyashris|away | 14:47 | |
| tobiash | mordred: if you have a sec for log annotation patches: https://review.opendev.org/#/q/topic:log-annotation+(status:open+OR+status:merged) | 15:07 |
| *** jcapitao has quit IRC | 15:17 | |
| *** nils has quit IRC | 15:28 | |
| *** sgw1 has joined #zuul | 15:29 | |
| openstackgerrit | Tobias Henkel proposed zuul/zuul master: Make reporting asynchronous https://review.opendev.org/691253 | 15:37 |
| *** sshnaidm|mtg is now known as sshnaidm | 15:39 | |
| corvus | mordred: a simple test fix: https://review.opendev.org/743628 | 15:40 |
| tobiash | mordred: thanks, could you +3 this test fix? https://review.opendev.org/743628 | 15:40 |
| corvus | tobiash: jinx :) | 15:41 |
| tobiash | :) | 15:41 |
| tobiash | corvus: would you be opposed to make reporting async? (I know that this is no problem with ha scheduler anymore, but that is months away and we are more and more backlogged in our event handling) | 15:42 |
| tobiash | we're now at 4.5k jobs per hour which means that we kick off more than one job per second and I see frequent multi-second delays in the event loop which adds up | 15:44 |
| corvus | tobiash: i think it's probably okay; my main worry is with adding complexity here, and therefore creating new emergent behaviors (new race condition bugs, etc). give me a little while to look over your change and really think about it. | 15:47 |
| tobiash | corvus: I just rebased the change so it's likely not working/complete yet | 15:48 |
| tobiash | I just wanted to clarify if I can continue working on it | 15:48 |
| corvus | tobiash: ack, thx. i'll review it in that context | 15:49 |
| tobiash | and I agree with the worry about the added complexity at I'd be in favor of removing it with ha scheduler again if that helps arguing it :) | 15:49 |
| corvus | tobiash: heh, my first thought is that if, by the time we have distributed schedulers, it's been in production for a while, we might just keep it :) | 15:50 |
| tobiash | with ha scheduler removing it likely means reducing zk traffic a bit :) | 15:51 |
| corvus | true; depends on exactly how we end up doing that | 15:52 |
| tobiash | a second thing I found that is potentially slowing down event processing is this one: https://opendev.org/zuul/zuul/src/branch/master/zuul/manager/independent.py#L78 | 15:54 |
| tobiash | this does also web requests and I wonder if we can do this work already ahead of time before enqueuing the change | 15:54 |
| *** hamalq has joined #zuul | 16:03 | |
| *** hamalq has quit IRC | 16:03 | |
| openstackgerrit | Clark Boylan proposed zuul/nodepool master: Add basic max-servers handling to the k8s driver https://review.opendev.org/743424 | 16:03 |
| *** hamalq has joined #zuul | 16:04 | |
| clarkb | y2kenny isn't here but I think ^ is a decent simplification (thank you corvus for that simple driver interface) | 16:09 |
| clarkb | basically we can add features and delete some code at the same time :) | 16:09 |
| tobiash | clarkb: I think at a later point we could also move launchesComplete into the driver interface | 16:12 |
| tobiash | that actually is the same for all drivers | 16:12 |
| clarkb | ++ | 16:12 |
| tobiash | *almost | 16:12 |
| *** wuchunyang has joined #zuul | 16:32 | |
| *** wuchunyang has quit IRC | 16:37 | |
| openstackgerrit | Merged zuul/zuul master: Annotate getChangeByUrl logs with event id https://review.opendev.org/743971 | 16:57 |
| openstackgerrit | Merged zuul/zuul master: Annotate all canMerge calls with event id https://review.opendev.org/743977 | 16:57 |
| openstackgerrit | Merged zuul/zuul master: Sequence builds in test_executor https://review.opendev.org/743628 | 16:57 |
| *** rlandy has joined #zuul | 17:02 | |
| *** jpena is now known as jpena|off | 17:04 | |
| *** sanjayu__ has quit IRC | 17:53 | |
| *** hashar has quit IRC | 17:54 | |
| *** bhavikdbavishi has quit IRC | 17:59 | |
| *** bhavikdbavishi has joined #zuul | 17:59 | |
| openstackgerrit | Andrii Ostapenko proposed zuul/zuul-jobs master: Add ability to use (upload|promote)-docker-image roles in periodic jobs https://review.opendev.org/740560 | 18:42 |
| *** bhavikdbavishi has quit IRC | 18:53 | |
| clarkb | swest: I left a comment on https://review.opendev.org/#/c/738013/5 can you let me know if that makes sense? Unfortuantely I don't have a great suggestion for fixing the underlying issue | 19:04 |
| clarkb | bschanzel: tristanC you can probably answer my question on https://review.opendev.org/#/c/740691/3 ? | 19:11 |
| *** y2kenny has joined #zuul | 19:23 | |
| *** y2kenny49 has joined #zuul | 19:25 | |
| *** y2kenny78 has joined #zuul | 19:26 | |
| *** y2kenny has quit IRC | 19:28 | |
| *** y2kenny has joined #zuul | 19:30 | |
| tristanC | clarkb: replied, thanks | 19:30 |
| clarkb | tristanC: '{"ENV-NAME": "value"}' is valid json :) | 19:33 |
| corvus | i think {'foo-bar': baz} is fine in yaml; but if that's what k8s is expecting, it probably makes the most sense to pass it through | 19:33 |
| corvus | in fact i think "{foo-bar: baz}" is valid yaml | 19:33 |
| y2kenny | clarkb: I saw the change. I will pull it down and try the test again. Is the simple driver thing something that should be adopted by most driver? (I am wondering if I should apply that to the cobbler driver... I wrote it referencing the k8s driver so I wasn't using it.) | 19:33 |
| y2kenny | unrelated: I have a job that is known to be flicky but it's ok as long as it passes once. I noticed that there is job.attempts but that seems to apply only to pre-run. Is the strategy to move the main-run to pre correct? Are there other gotcha that I should pay attention to? | 19:34 |
| clarkb | y2kenny: I expect that over time most drivers will adopt it or something like it as a base | 19:34 |
| corvus | y2kenny: (yes, and i don't think it existed at the time you started your work) | 19:34 |
| clarkb | y2kenny: since as we write more of these we are finding more overlap (the quota stuff is a good example of that) | 19:34 |
| corvus | y2kenny: can you put the test itself in a retry loop? | 19:35 |
| corvus | y2kenny: that way you don't have to re-run the whole job | 19:35 |
| clarkb | firefox just crashed, but I'm just doing quick rereview of that k8s change and will approve it likely (given that k8s expects that data structure) | 19:36 |
| clarkb | my laptop is getting very flaky, I should probably run memcheck on it | 19:37 |
| clarkb | :/ | 19:37 |
| corvus | y2kenny: (but just as an aside, from experience: flaky tests add up *fast*, so you can pretty quickly end up with a useless test suite if you don't stay on top of them) | 19:37 |
| y2kenny | corvus: might be difficult for this particular job. The job launches a VM via kubevirt and it doesn't fail very cleanly so I was just thinking of relaunching the VM. | 19:37 |
| y2kenny | corvus: it's not really a test... the job pretty much just launches a VM and install a few package to build a PXE image. It should really never fail but once in a while it seems to run into some network stability problem. | 19:38 |
| corvus | y2kenny: then putting everything in pre would be a quick fix, especially if you aren't concerned with the result | 19:39 |
| tristanC | clarkb: "ENV-NAME: value" is also valid yaml, but it seems safer to put arbritary key name as a value. | 19:39 |
| corvus | tristanC: i don't think it's more or less safe | 19:40 |
| corvus | tristanC: but if it's what k8s expects, then i think it makes sense to pass it through unchanged | 19:40 |
| clarkb | I approved the change since that is what k8s expects and keeping the same structure will be useful for users to not map | 19:40 |
| clarkb | but I also think its silly k8s does that :) | 19:40 |
| y2kenny | corvus: so in the kickstart/install step it just couldn't dnf install something... I should chase down issue but it's hard to say where the instability come from at this point (it could be kubevirt, k8s networking, my network environment or the network provided by my IT.) | 19:41 |
| corvus | y2kenny: always start by blaming the network; the networking folks like that :) | 19:42 |
| fungi | even better, blame the firewall. sometimes that's even different folks still | 19:43 |
| tristanC | corvus: clarkb: i meant safer in the sense of less ambiguous. yaml does all sorts of thing that are not obvious, such as triming trailing space on keys | 19:44 |
| y2kenny | fungi: oh yes that actually happens a lot... and the funny part is the network guys blame things on the firewall/infosec guys :) | 19:44 |
| clarkb | tristanC: >>> yaml.dump({' foo ': ' bar '}) -> "' foo ': ' bar '\n" | 19:46 |
| y2kenny | clarkb: ok I will look through that and adapt what I can. (I plan to upstream the cobbler driver btw... it's just that I still need to get the high-ups to approve open source contribution. And would be easier if I can demonstrate the value of my infrastructure.) | 19:46 |
| tristanC | clarkb: right, so you need to quote the keys. | 19:47 |
| tristanC | clarkb: if your user typed " foo : bar" -> {'foo': 'bar'} | 19:48 |
| clarkb | tristanC: right but that isn't safer because its still true of the value side | 19:48 |
| corvus | i recognize people may have different perspectives on this, but i find it not at all surprising that one would have to quote keys if one thought whitespace were significant | 19:48 |
| clarkb | basically you can still get it wrong doing things the k8s way | 19:49 |
| corvus | and yes, the same is true for values | 19:49 |
| openstackgerrit | Merged zuul/nodepool master: OpenShift/k8s Provider: Allow passing env vars to Pods https://review.opendev.org/740691 | 20:17 |
| y2kenny | clarkb: confirmed 743424/8 still works. | 20:24 |
| clarkb | y2kenny: great. I imagine we justneed reviews now | 20:27 |
| corvus | clarkb: lgtm; do you think we can/should add any more testing for that? | 20:35 |
| clarkb | corvus: ya maybe we should add a second case to the existing test setup that tries to start 3 pods | 20:37 |
| clarkb | I'm finishing lunch but let me see how difficult that would be | 20:38 |
| clarkb | oh I was thinking of it from the functional test job perspective but I bet its even easier in the unittests | 20:39 |
| *** vishalmanchanda has quit IRC | 20:39 | |
| corvus | clarkb: yeah, that sounds useful; iiuc, the existing case at least exercises the code, but seeing the cap work would be nice | 20:39 |
| *** ianychoi has joined #zuul | 20:41 | |
| openstackgerrit | Clark Boylan proposed zuul/nodepool master: Add basic max-servers handling to the k8s driver https://review.opendev.org/743424 | 21:04 |
| clarkb | corvus: ^ that adds a unittest | 21:04 |
| clarkb | heh and I fail pep8, sorry new ps in a moment | 21:04 |
| openstackgerrit | Clark Boylan proposed zuul/nodepool master: Add basic max-servers handling to the k8s driver https://review.opendev.org/743424 | 21:05 |
| *** sshnaidm is now known as sshnaidm|afk | 21:42 | |
| *** rlandy is now known as rlandy|biab | 23:13 | |
| *** tosky has quit IRC | 23:14 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!