| *** tosky has quit IRC | 00:22 | |
| *** jamesmcarthur has joined #zuul | 00:35 | |
| *** mhu has quit IRC | 00:46 | |
| mnaser | ianw: you know what i think ive seen thatttttt | 00:48 |
|---|---|---|
| mnaser | i remember tracking it down to the exact spot | 00:49 |
| mnaser | let me try to find it | 00:49 |
| mnaser | ianw: i think disabling checksumming actually made it pass this https://github.com/openstack/diskimage-builder/blob/cfba9ea79d903574afe58bbe3964c050676ee8da/diskimage_builder/lib/common-functions#L68-L72 | 00:50 |
| mnaser | im not sure if i fully understood that part | 00:50 |
| ianw | mnaser: yeah, that's exactly the point | 01:15 |
| ianw | the only thing i think it can be waiting for is the outfilter.py that we run things through | 01:16 |
| ianw | we could just wait for those pids ... i don't know if that is a fix or a horrible hack papering over the real problem | 01:18 |
| mnaser | ianw: i remember looking into it and the outfilter having to do with it i think | 01:23 |
| mnaser | im sorry i cant do much other than "i confirm what you're seeing" heh | 01:23 |
| ianw | no that's super useful! nice to know i'm not alone :) | 01:23 |
| *** bhavikdbavishi has joined #zuul | 02:57 | |
| *** bhavikdbavishi has quit IRC | 02:59 | |
| *** jamesmcarthur has quit IRC | 03:09 | |
| *** jamesmcarthur has joined #zuul | 03:10 | |
| *** jamesmcarthur has quit IRC | 03:15 | |
| ianw | i swear there's some sort of race condition with https://review.opendev.org/#/c/693464/20 | 03:19 |
| *** rlandy|rover|bbl has quit IRC | 03:20 | |
| ianw | twice it reported "This change depends on a change with an invalid configuration." | 03:20 |
| ianw | then nothing changed, but the 3rd recheck and it started | 03:20 |
| ianw | https://paste.fedoraproject.org/paste/xnxogv52-DnCiJbc3CjMrA | 03:23 |
| ianw | it looks like it's https://review.opendev.org/#/c/694177/ in the chain it doesn't like | 03:25 |
| ianw | that Depends-On: https://review.opendev.org/694175 ... which is a dib change ... this is cross-tenant | 03:25 |
| *** jamesmcarthur has joined #zuul | 03:40 | |
| *** jamesmcarthur has quit IRC | 03:45 | |
| *** jamesmcarthur has joined #zuul | 04:01 | |
| *** jamesmcarthur has quit IRC | 04:14 | |
| *** jamesmcarthur has joined #zuul | 04:15 | |
| *** jamesmcarthur has quit IRC | 04:21 | |
| *** jamesmcarthur has joined #zuul | 04:46 | |
| *** jamesmcarthur has quit IRC | 04:52 | |
| *** bjackman has joined #zuul | 05:04 | |
| *** raukadah is now known as chandankumar | 05:09 | |
| *** igordc has quit IRC | 05:11 | |
| ianw | https://zuul.opendev.org/t/zuul/build/9f38348bffaa43799c807e055e070974 ... it worked! nodepool-launcher & builder containers in the functional job | 05:26 |
| *** reiterative has quit IRC | 05:35 | |
| *** swest has joined #zuul | 06:25 | |
| *** jamesmcarthur has joined #zuul | 06:48 | |
| *** jamesmcarthur has quit IRC | 06:52 | |
| *** saneax has joined #zuul | 07:01 | |
| *** bjackman has quit IRC | 07:08 | |
| *** chandankumar is now known as chkumar|rover | 07:22 | |
| *** pcaruana has joined #zuul | 07:27 | |
| *** jamesmcarthur has joined #zuul | 07:28 | |
| *** tosky has joined #zuul | 07:31 | |
| *** jamesmcarthur has quit IRC | 07:34 | |
| *** bolg has joined #zuul | 07:48 | |
| *** avass is now known as Guest72500 | 07:51 | |
| *** avass has joined #zuul | 07:51 | |
| *** adso78 has joined #zuul | 07:55 | |
| *** bolg has quit IRC | 08:02 | |
| *** bolg has joined #zuul | 08:03 | |
| *** bolg has quit IRC | 08:05 | |
| *** bolg has joined #zuul | 08:15 | |
| *** jamesmcarthur has joined #zuul | 08:15 | |
| *** bolg has quit IRC | 08:17 | |
| *** jamesmcarthur has quit IRC | 08:20 | |
| *** bolg has joined #zuul | 08:27 | |
| *** bolg has quit IRC | 08:29 | |
| *** jpena|off is now known as jpena | 08:35 | |
| *** reiterative has joined #zuul | 08:38 | |
| *** bolg has joined #zuul | 08:39 | |
| *** bolg has quit IRC | 08:41 | |
| *** jamesmcarthur has joined #zuul | 08:47 | |
| *** bolg has joined #zuul | 08:51 | |
| *** jamesmcarthur has quit IRC | 08:52 | |
| *** bolg has quit IRC | 08:55 | |
| *** bolg has joined #zuul | 09:02 | |
| *** bolg has quit IRC | 09:07 | |
| *** bolg has joined #zuul | 09:14 | |
| *** sshnaidm|ruck is now known as sshnaidm | 09:19 | |
| *** bolg has quit IRC | 09:20 | |
| *** bolg has joined #zuul | 09:26 | |
| *** bolg has quit IRC | 09:30 | |
| *** saneax has quit IRC | 09:44 | |
| *** saneax has joined #zuul | 09:44 | |
| *** sshnaidm has quit IRC | 10:01 | |
| *** sshnaidm has joined #zuul | 10:05 | |
| *** jamesmcarthur has joined #zuul | 10:49 | |
| *** jamesmcarthur has quit IRC | 10:53 | |
| *** rfolco has joined #zuul | 11:18 | |
| *** rfolco is now known as rfolco|doctor | 11:18 | |
| *** jamesmcarthur has joined #zuul | 11:25 | |
| *** jamesmcarthur has quit IRC | 11:31 | |
| *** avass has quit IRC | 12:00 | |
| *** jamesmcarthur has joined #zuul | 12:01 | |
| *** jpena is now known as jpena|lunch | 12:14 | |
| *** jamesmcarthur has quit IRC | 12:50 | |
| *** jamesmcarthur_ has joined #zuul | 12:50 | |
| *** rlandy has joined #zuul | 12:58 | |
| *** jpena|lunch is now known as jpena | 13:20 | |
| *** rfolco|doctor is now known as rfolco | 13:23 | |
| fungi | zuul-maint: an interesting problem... right now https://zuul.opendev.org/api/tenant/openstack/status has well over 1k entries in its "gate' pipeline, almost all of which have an empty heads list | 13:26 |
| fungi | if the dashboard attempts to render that content, it kills your browser (or your entire computer) | 13:27 |
| fungi | i saved a snapshot since the last time folks reported similar browser symptoms they cleared up on their own eventually | 13:28 |
| *** hashar has joined #zuul | 13:28 | |
| fungi | er, now that i think about it, was the builds dashboard chewing up folks' browsers last week not status | 13:29 |
| AJaeger | fungi: http://zuul.opendev.org/t/openstack/jobs kills my browser - is that calling to /status? | 13:29 |
| fungi | oh, wow. i wouldn't think so | 13:30 |
| *** hashar has quit IRC | 13:32 | |
| *** yolanda has quit IRC | 13:33 | |
| *** yolanda has joined #zuul | 13:36 | |
| tobiash | fungi: interesting, I can observe the same thing in our deployment | 13:38 |
| tobiash | thanks for the note | 13:38 |
| tobiash | that might even explain the memleak we still see in the scheduler... | 13:39 |
| tobiash | AJaeger: we have a similar problem with the jobs page, there the js seems to oom the browser | 13:41 |
| tobiash | but the jobs page is a different problem | 13:41 |
| AJaeger | tobiash: the jobs page is the problem, not status... | 13:42 |
| tobiash | AJaeger: fungi wrote about the status endpoint (which is used by the status page) | 13:42 |
| tobiash | so that are actually two problems | 13:42 |
| tobiash | * the status endpoint seems to contain old entries | 13:43 |
| tobiash | * the jobs page ooms the browser with many jobs or maybe jobs with special config | 13:44 |
| *** mhu has joined #zuul | 13:45 | |
| fungi | yeah, sorry, trying to debug it in here and in #openstack-infra at the same time | 13:46 |
| tobiash | ok, looking more closely at the status endpoint it looks like the dependent pipeline doesn't delete empty change queues -> all earlier change queues are reported | 13:46 |
| fungi | i misunderstood the initial problem report and thought it was /status but it's /jobs | 13:46 |
| tobiash | that makes one per repo that was in use since the last scheduler startup | 13:46 |
| tobiash | ah ok, we see the jobs problem as well in one of our bigger tenants | 13:47 |
| tristanC | adding some console.log in my local copy, it seems like the jobs page is having issue with the one named "opendev-uploader-docker-image" | 13:48 |
| tobiash | tristanC: so it might be a special job construct and not just the number of jobs? | 13:48 |
| fungi | yeah, i doubt it's number of jobs | 13:49 |
| tristanC | tobiash: i can't tell, reloading the page make the debug stop at another job name | 13:50 |
| tobiash | hrm | 13:50 |
| fungi | i watched a "web content" subprocess in firefox 70 consume ~11gb of memory in a matter of a few minutes trying to load the jobs dashboard | 13:50 |
| mhu | hey there, could I get reviews on https://review.opendev.org/#/c/641099/ ? it's been in limbo for a while | 13:50 |
| fungi | so seems like something self-referential maybe causing runaway memory allocation | 13:50 |
| fungi | infinite loop kinda territory | 13:50 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: authentication config: add optional token_expiry https://review.opendev.org/642408 | 13:51 |
| tristanC | oh wait no, the console wasn't scrolled after my refresh, it did stopped again at opendev-uploader-docker-image | 13:51 |
| tobiash | oh cool, now the question is, is there anything special with this job? | 13:51 |
| tristanC | I added a console.log("getNode(", job, ")") here https://opendev.org/zuul/zuul/src/branch/master/web/src/containers/jobs/Jobs.jsx#L63 | 13:52 |
| fungi | huh, http://zuul.opendev.org/t/openstack/job/opendev-uploader-docker-image returns a 404 for http://zuul.opendev.org/api/tenant/openstack/job/opendev-uploader-docker-image | 13:53 |
| tobiash | I don't see it in the jobs endpoint as well | 13:54 |
| fungi | did you mean opendev-upload-docker-image maybe? | 13:55 |
| fungi | http://zuul.opendev.org/t/openstack/job/opendev-upload-docker-image returns details | 13:55 |
| fungi | the parentage on it seems fairly straightforward too, if that's the one | 13:55 |
| fungi | and there are no builds for that job | 13:57 |
| tobiash | could it be a problem that this job is listed as parent before it's listed itself? http://paste.openstack.org/show/786491/ | 13:57 |
| fungi | though a side note, it might be nice if the job description renderer could handle sphinx include tags | 13:57 |
| tristanC | fungi: yeah, though i think it's the api that should resolve the include directive | 13:59 |
| fungi | tristanC: ahh, good point, the dashboard lacks sufficient context to find those inclusions | 14:00 |
| fungi | tobiash: i would be surprised if the dashboard code relied on jobs being sorted in ascending order of ancestry | 14:01 |
| fungi | i would expect the problem to be far more prevalent since i doubt we intentionally sort that data (though i certainly could be mistaken) | 14:01 |
| tristanC | tobiash: it should'nt matter, the list is processed in multiple stage | 14:01 |
| tobiash | ok | 14:01 |
| tristanC | i'm commenting parent resolution code trying to find what could be the issue, so far no luck | 14:02 |
| AJaeger | accessing /jobs worked a few weeks ago - did we add so many more jobs or is this a regression from some other change? | 14:05 |
| tristanC | commenting this block https://opendev.org/zuul/zuul/src/branch/master/web/src/containers/jobs/Jobs.jsx#L66-L73 makes the ui work | 14:07 |
| *** Goneri has quit IRC | 14:09 | |
| *** jamesmcarthur_ has quit IRC | 14:10 | |
| fungi | that loop does seem ripe for recursion issues | 14:11 |
| fungi | though i don't immediately see any self-referential calls there | 14:13 |
| tristanC | fungi: the parents list is processed later down the loop | 14:14 |
| tristanC | the issue start with this job zuul-upload-image (when adding a console.log here: https://opendev.org/zuul/zuul/src/branch/master/web/src/containers/jobs/Jobs.jsx#L127 | 14:15 |
| fungi | aha, so you were effectively zeroing out the parents list | 14:16 |
| fungi | causing the processing loop to no-op | 14:17 |
| *** jamesmcarthur has joined #zuul | 14:20 | |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: enqueue: make trigger optional https://review.opendev.org/695446 | 14:28 |
| tristanC | so the getNode doesn't seems to be the issue, i think it's the tree view component that is getting confused by a bad parents list, e.g. the page also works by commenting https://opendev.org/zuul/zuul/src/branch/master/web/src/containers/jobs/Jobs.jsx#L142 | 14:30 |
| *** tosky_ has joined #zuul | 14:31 | |
| tristanC | thus i don't think it's related to the zuul-upload-image or opendev-upload-docker-image, it could be caused by another job | 14:32 |
| *** tosky has quit IRC | 14:32 | |
| *** tosky_ is now known as tosky | 14:33 | |
| tristanC | preventing adding a jobnode to multiple parent also fixes the issue, here is another fix and a list of suspicious job: http://paste.openstack.org/show/786495/ | 14:39 |
| corvus | so the issue is that the tree view, naturally, can't display nodes with multiple parents | 14:42 |
| *** fdegir has quit IRC | 14:42 | |
| fungi | makes sense | 14:43 |
| corvus | given that, this is always going to be an approximation, so i think that fix (use the parent of the first variant) is probably best | 14:43 |
| *** fdegir has joined #zuul | 14:43 | |
| tristanC | corvus: iirc it used to work, we made that whole get each parent per variant to support the use-case where a job can inherit from multiple parent... | 14:43 |
| corvus | tristanC: oh, did we show the job multiple times in the tree? | 14:44 |
| corvus | that's the only other way i can think to handle this | 14:44 |
| tristanC | corvus: i thought we did, perhaps i don't remember correctly. | 14:44 |
| fungi | i suppose if we wanted to get fancy we could add some sort of indicator to nodes with additional parents (could even set an on-click handler to jump/highlight additional enumerated parents, though i doubt that would be particularly useful) | 14:46 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 14:48 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul master: web: only display the first parent in the jobs tree https://review.opendev.org/695450 | 14:49 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: enqueue: make trigger optional https://review.opendev.org/695446 | 14:51 |
| tristanC | corvus: in case multi parent never worked, here is the easy fix ^ | 14:51 |
| corvus | tristanC: i'm happy to +2 that (and that seems like a good quick fix for the current breakage), but also happy to support rendering the job multiple times if you prefer that as a followup. thanks for digging in. :) | 14:51 |
| tristanC | corvus: yeah, i think it may have worked in simple case, but perhaps when the schema is more complex (e.g. a job attached to a multi-parented parent) then the component is unable to render the tree | 14:54 |
| *** tosky has quit IRC | 14:54 | |
| tristanC | that could be fixed by using unique job's React.Fragment per attached parent | 14:56 |
| tristanC | i can have a look later | 14:56 |
| *** tosky has joined #zuul | 14:57 | |
| *** igordc has joined #zuul | 15:01 | |
| *** pcaruana has quit IRC | 15:02 | |
| *** igordc has quit IRC | 15:05 | |
| *** igordc has joined #zuul | 15:06 | |
| *** igordc has quit IRC | 15:14 | |
| *** igordc has joined #zuul | 15:14 | |
| *** pcaruana has joined #zuul | 15:20 | |
| AJaeger | tristanC: thanks for the fix! | 15:25 |
| *** jpena is now known as jpena|off | 15:41 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 15:47 |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul master: web: handle jobs that have multiple parent https://review.opendev.org/695450 | 15:55 |
| tristanC | corvus: i think this is a better fix, but i'm not sure how to differentiate the branch based variants from the multi-parent ones | 15:56 |
| tristanC | corvus: actually i can't find the example where two job variants inherit from different parent. iirc it was used by some docker-build? | 16:04 |
| *** chkumar|rover is now known as raukadah | 16:08 | |
| tristanC | oh found it, https://review.opendev.org/#/c/629983/3/.zuul.yaml . So if that's something we want to support, shouldn't the parent attribute be a list, so that we only have variants per branch? | 16:09 |
| tobiash | tristanC: we already discussed this a while ago, so atm we accidentally support multiple inheritance by this process and I think nobody objected making this official by making the parents a list | 16:10 |
| tobiash | but I think nobody has implemented it yet | 16:10 |
| tristanC | tobiash: it seems like that would make the model simpler, then 'variants' would be equivalent to 'branches' iiuc | 16:11 |
| openstackgerrit | Matthieu Huin proposed zuul/zuul master: [WIP] admin REST API: zuul-web integration https://review.opendev.org/643536 | 16:20 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 16:26 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 16:54 |
| *** jamesmcarthur has quit IRC | 17:27 | |
| *** jpena|off is now known as jpena | 17:27 | |
| *** jamesmcarthur has joined #zuul | 17:28 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support. https://review.opendev.org/695051 | 17:41 |
| *** michael-beaver has joined #zuul | 17:50 | |
| *** pcaruana has quit IRC | 17:51 | |
| *** mattw4 has joined #zuul | 17:51 | |
| openstackgerrit | Tristan Cacqueray proposed zuul/zuul master: config: add tenant.toDict() method and REST endpoint https://review.opendev.org/621344 | 18:07 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: install-podman: also install uidmap https://review.opendev.org/695545 | 18:09 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 18:10 |
| *** pcaruana has joined #zuul | 18:14 | |
| *** tosky has quit IRC | 18:29 | |
| *** gmann is now known as gmann_afk | 18:38 | |
| *** saneax has quit IRC | 18:42 | |
| *** pcaruana has quit IRC | 18:42 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 18:51 |
| *** pcaruana has joined #zuul | 19:06 | |
| corvus | oh sweet, that patch has gotten to the point where the errors are now actually in the 'run-buildset-registry' role :) | 19:25 |
| fungi | progressive error development | 19:26 |
| mordred | \o/ | 19:29 |
| *** jpena is now known as jpena|off | 19:36 | |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 19:37 |
| *** openstackstatus has quit IRC | 19:50 | |
| openstackgerrit | Merged zuul/zuul master: Zuul Web: add /api/user/authorizations endpoint https://review.opendev.org/641099 | 19:50 |
| *** dtroyer has quit IRC | 19:50 | |
| *** dtroyer has joined #zuul | 19:51 | |
| *** openstackstatus has joined #zuul | 19:52 | |
| *** ChanServ sets mode: +v openstackstatus | 19:52 | |
| *** mattw4 has quit IRC | 19:54 | |
| *** mattw4 has joined #zuul | 19:54 | |
| *** ianw has quit IRC | 20:20 | |
| *** ssbarnea has quit IRC | 20:26 | |
| *** ianw has joined #zuul | 20:27 | |
| openstackgerrit | Merged zuul/zuul-jobs master: install-podman: also install uidmap https://review.opendev.org/695545 | 20:39 |
| openstackgerrit | James E. Blair proposed zuul/zuul-jobs master: WIP: use-buildset-registry: add podman support https://review.opendev.org/695051 | 20:39 |
| ianw | corvus / modred: i got the nodepool functional job working with containerised builder & launcher yesterday ... which means as i start the process of getting it all reviewable i'll have to make a decision on the container namespace for dib | 20:42 |
| ianw | https://review.opendev.org/#/c/693971/ | 20:43 |
| ianw | i don't really mind, but will have to choose something :) | 20:43 |
| ianw | it feels like opendev/diskimage-builder works? but not moving it to opendev/ in git, to avoid making work for people | 20:44 |
| *** gmann_afk is now known as gmann | 20:45 | |
| corvus | ianw: this is probably more of an infra question | 20:45 |
| ianw | corvus: ok, so we can probably rule out zuul/diskimage-builder? | 20:46 |
| corvus | oh yeah, sorry. i think so. i think we considered that when we started the zuul project and decided it wasn't the right place. i don't think that has changed. so i think it's mostly an openstack/opendev question. | 20:47 |
| ianw | ok; and from a nodepool point of view, inheriting from either doesn't really make a difference? | 20:48 |
| corvus | ianw: not that i can think of. now, part of me thinks it's weird for it to inherit from dib at all (since dib is one component of one part of one driver). it seems like it should just be adding in what's needed for dib to work. | 20:50 |
| corvus | like, what if we needed something else to build aws images, would we inherit from that? | 20:50 |
| corvus | but i haven't thought about this much | 20:50 |
| *** rfolco has quit IRC | 20:50 | |
| clarkb | personally I think it should be ok to produce an image that works with many drivers | 20:51 |
| corvus | but that seems easy to change later; i don't think getting the perfect inheritance ordering needs to block anything. | 20:51 |
| clarkb | optimizing in this particular space seems to always cause headaches for users when they realize they cant use $oyherthing because it isnt supported in the current special builf | 20:51 |
| corvus | it looks like the dockerfile for the dib image is just installing dib; that seems like something we could put in the dockerfile for nodepool-builder | 20:52 |
| ianw | i think the reason to do that was that it was installing all the dependencies, etc; stuff that we wanted isolated from the builder image | 20:53 |
| ianw | like if dib has a new dependency, you don't have to update the builder image because it's layer on | 20:53 |
| corvus | makes sense. that works for this case, but as soon as we have a second thing like that, we're back in the same boat. so it's probably unsafe to consider it permanent. | 20:54 |
| ianw | mordred did actually start more or less that suggestion (add dib to nodepool) https://review.opendev.org/#/c/693306/ | 20:54 |
| ianw | we already sort of do i guess - openstacksdk? we want to add that into the mix for testing too | 20:55 |
| corvus | i wonder if we could run bindep on nodepool's bindep file | 20:55 |
| corvus | er dib's bindep file | 20:55 |
| ianw | that would mean the python builder was cloning dib source? | 20:57 |
| corvus | yeah, or otherwise getting the source (or that file) into the build image. | 20:58 |
| corvus | (i'm flexible when it comes to dockerfile purity :) | 20:58 |
| mordred | corvus: I thought a bit about doing that - but it seemed to me like we'd essentially be building something very similar to the dib image in the first phase of such a nodepool dockerfile - so why not just use the dib image as a base. I agree, it's a little weird :) | 20:59 |
| *** openstack has joined #zuul | 21:18 | |
| *** ChanServ sets mode: +o openstack | 21:18 | |
| *** sshnaidm has quit IRC | 21:21 | |
| *** pcaruana has quit IRC | 21:25 | |
| clarkb | also re the specific example of aws I believe that dib is probably 95% capable of producing ec2 images already. There is a good chance that it might even work already? | 21:28 |
| corvus | yeah, i was just trying to come up with an example | 21:31 |
| corvus | the specifics aren't important; what's important is that we explicitly designed the architecture so that dib isn't required. it's just one implementation of one (could be more later) optional part of the system. | 21:32 |
| corvus | it should be part of the nodepool-builder image, but it shouldn't always be the only thing on there. | 21:33 |
| clarkb | yup I agree. I just didn't want people thinking for aws they had to use something other than dib | 21:33 |
| corvus | ++ | 21:33 |
| clarkb | looks like the aws image import tool (you upload image to s3 then run import-image against that object) supports raw images. In this case dib should be fine as is. Just need nodepool-builder bits that know how to do that process against aws | 21:34 |
| corvus | it's sort of looking like if i start a container with podman as a regular user, it doesn't show up in "podman ps" run as that user, but does with sudo | 21:34 |
| corvus | i don't understand that | 21:34 |
| corvus | that's not the behavior i see locally, but i do see that in a zuul job | 21:35 |
| clarkb | could it be that the ps process namespace is disjoint from the podman container process namespace? | 21:36 |
| corvus | in fact, locally, i don't see a user container when i run the command as root | 21:36 |
| clarkb | I would expect that could happen if both are running in containers, but if running in the root process namespace I would expect what you describe locally | 21:36 |
| corvus | oh, i see it | 21:37 |
| corvus | it was started as root; i just missed it because the "become: true" was like taskfile levels up | 21:37 |
| corvus | it would be cool if the become flag trickled down to the leaf node json data, so that the root tasks always showed that regardless of where it was set | 21:38 |
| *** tjgresha has quit IRC | 22:12 | |
| *** sshnaidm has joined #zuul | 22:17 | |
| corvus | remote: https://review.opendev.org/695601 install-podman: also install slirp4netns | 22:29 |
| *** tosky has joined #zuul | 22:35 | |
| *** lennyb has joined #zuul | 22:41 | |
| *** mhu has quit IRC | 23:34 | |
| *** jamesmcarthur has quit IRC | 23:52 | |
| *** jamesmcarthur has joined #zuul | 23:53 | |
| Shrews | slirp? wow, such memories of hacking up college internet access | 23:53 |
| *** jamesmcarthur has quit IRC | 23:57 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!