| *** JasonCL has quit IRC | 00:00 | |
| *** JasonCL has joined #zuul | 00:39 | |
| *** odyssey4me has quit IRC | 00:47 | |
| *** JasonCL has quit IRC | 00:47 | |
| *** odyssey4me has joined #zuul | 00:47 | |
| *** JasonCL has joined #zuul | 01:13 | |
| *** harlowja has quit IRC | 01:14 | |
| dmsimard | Had never looked at what GitLab CI was | 01:25 |
|---|---|---|
| dmsimard | I'll stay with Zuul | 01:25 |
| dmsimard | https://docs.gitlab.com/ee/ci/yaml/ | 01:25 |
| *** JasonCL has quit IRC | 01:41 | |
| *** JasonCL has joined #zuul | 01:49 | |
| tristanC | bitbucket also comes with it's own integrated ci: https://bitbucket.org/product/features/pipelines | 01:49 |
| *** JasonCL has quit IRC | 01:55 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: mqtt: add basic reporter https://review.openstack.org/535543 | 02:09 |
| *** bhavik1 has joined #zuul | 02:13 | |
| *** bhavik1 has quit IRC | 02:16 | |
| *** Wei_Liu has joined #zuul | 02:39 | |
| mrhillsman | the default roles available for uploading logs does not allow changing the ssh port true? | 03:04 |
| mrhillsman | s/roles/role | 03:04 |
| mrhillsman | personally i do not think it does but am not 100% sure | 03:04 |
| clarkb | mrhillsman: ya doesnt look like add-fileserver role takes aport argument | 03:13 |
| mrhillsman | ok thx | 03:13 |
| clarkb | you could likely extend it pretty simply especially since it alrwady takes a bundle of data in the secret | 03:14 |
| mrhillsman | ++ | 03:14 |
| mrhillsman | ty sir | 03:14 |
| tobiash | corvus: re 535680: it works and we use it in production since a few months | 04:16 |
| tobiash | corvus: basically it's WIP because I didn't have time yet to add tests to it | 04:17 |
| *** harlowja has joined #zuul | 04:23 | |
| *** rlandy has quit IRC | 04:31 | |
| *** swest has joined #zuul | 05:01 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Also prefix the alembic_version table https://review.openstack.org/560263 | 05:18 |
| tobiash | corvus, clarkb: just noticed that for correctly support table prefixes I missed an important fix ^ | 05:19 |
| tobiash | without that prefixed tables won't work correctly as soon as there are more than one deployments sharing the same db | 05:20 |
| tobiash | unfortunately I cannot think of a way to fix that (if someone is already using prefixes) without manual intervention (aka renaming the version table by hand before upgrading) | 05:22 |
| *** elyezer has quit IRC | 05:25 | |
| *** elyezer has joined #zuul | 05:29 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Also prefix the alembic_version table https://review.openstack.org/560263 | 05:31 |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Add postgresql release note https://review.openstack.org/560079 | 05:33 |
| *** harlowja has quit IRC | 05:35 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: Add regex support to project stanzas https://review.openstack.org/535713 | 05:36 |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: DNM: test https://review.openstack.org/560265 | 05:41 |
| openstackgerrit | Merged openstack-infra/zuul master: Return CORS headers on all requests https://review.openstack.org/555027 | 05:56 |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: gerrit: recognize project-created event https://review.openstack.org/560274 | 06:14 |
| *** jhesketh_ is now known as jhesketh | 06:29 | |
| *** hashar has joined #zuul | 06:36 | |
| openstackgerrit | Tristan Cacqueray proposed openstack-infra/zuul master: mqtt: add basic reporter https://review.openstack.org/535543 | 06:49 |
| *** xinliang has quit IRC | 07:42 | |
| *** xinliang has joined #zuul | 07:42 | |
| *** jpena|off is now known as jpena | 07:55 | |
| *** electrofelix has joined #zuul | 08:40 | |
| *** hashar is now known as hasharAway | 09:48 | |
| *** Wei_Liu has quit IRC | 09:52 | |
| *** Wei_Liu has joined #zuul | 09:52 | |
| kklimonda | does Depends-On work in gate pipeline when for inter-project dependencies, where two projects don't share a queue? | 10:16 |
| tobiash | kklimonda: yes it should, but the second won't be enqueued automatically | 10:43 |
| kklimonda | tobiash: hmm, not sure what you mean | 10:44 |
| kklimonda | both will still have to have a correct set of votes to be enqueued, what I'm mostly interested in if both changes will be merged at the ~same time | 10:45 |
| tobiash | The gate will ensure that the dependency is there but they won't be enqueued into the gate at the same time | 10:45 |
| tobiash | But the order of merging is ensured | 10:46 |
| kklimonda | right, so there is no guarantee that the change A will be merged before change B (that depends on A) about the same time, but if change B gets its votes before change A, they should enter the gate around the same time | 10:48 |
| AJaeger | kklimonda: If b has "depends-on: A" then A will always merge first. That is guranteed | 11:05 |
| *** xinliang has quit IRC | 11:11 | |
| kklimonda | AJaeger: but whether A and B are merged together is dependent on whether they share a queue, right? | 11:15 |
| AJaeger | kklimonda: if they share a queue, the get merged one after the other (if they pass tests;) | 11:16 |
| *** xinliang has joined #zuul | 11:22 | |
| *** xinliang has quit IRC | 11:22 | |
| *** xinliang has joined #zuul | 11:22 | |
| *** jpena is now known as jpena|lunch | 11:46 | |
| *** hasharAway has quit IRC | 12:05 | |
| *** odyssey4me has quit IRC | 12:14 | |
| *** odyssey4me has joined #zuul | 12:14 | |
| *** rlandy has joined #zuul | 12:30 | |
| *** dkranz has joined #zuul | 12:36 | |
| *** jpena|lunch is now known as jpena | 12:50 | |
| *** _ari_|conf is now known as _ari_ | 13:17 | |
| *** JasonCL has joined #zuul | 13:22 | |
| *** iamcxl001 has joined #zuul | 13:26 | |
| *** iamcxl001 has left #zuul | 13:29 | |
| *** ttx has quit IRC | 13:55 | |
| *** ttx has joined #zuul | 13:57 | |
| corvus | mordred: can you +3 https://review.openstack.org/560090 pls? | 14:09 |
| *** dkranz has quit IRC | 14:34 | |
| openstackgerrit | Jeremy Stanley proposed openstack-infra/zuul master: Add instructions for reporting vulnerabilities https://review.openstack.org/554352 | 14:42 |
| fungi | corvus: ^ thanks for pointing out i should have updated that after we got the new docs site | 14:47 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul-jobs master: Add installation/deprecation documentation https://review.openstack.org/560463 | 15:04 |
| corvus | i encourage everyone to read that change, as it outlines a deprecation policy for zuul-jobs ^ | 15:05 |
| corvus | mordred, dmsimard, AJaeger, pabelanger, tristanC, tobiash: ^ | 15:06 |
| tobiash | hrm, got a strange exception on the executor: http://paste.openstack.org/show/718944/ | 15:17 |
| tobiash | I've never seen that before | 15:17 |
| tobiash | I just tried to add a hello world job with playbook | 15:17 |
| *** dkranz has joined #zuul | 15:22 | |
| corvus | tobiash: that seems to indicate that /tmp/eefa1daac14348c88cad40da7e1a0748/ansible/playbook_0/role_0 already exists | 15:42 |
| corvus | tobiash: are there any log lines before that which might suggest a symlink was already created there? | 15:42 |
| tobiash | I don't find anything in the logs | 15:45 |
| tobiash | the project containing that at least has a path roles in it | 15:46 |
| corvus | tobiash: running tip of master? | 15:46 |
| tobiash | but I have to retest with keep to check what's already symlinked there | 15:46 |
| tobiash | not yet | 15:46 |
| tobiash | 2-3 weeks old | 15:46 |
| corvus | tobiash: oh, so this doesn't have the role checkout changes | 15:46 |
| tobiash | so these might fix it | 15:47 |
| tobiash | ? | 15:47 |
| corvus | tobiash: no idea. there's a good chance the bug may still be present, but everything about that has changed, so it's probably worth upgrading before debugging further. | 15:47 |
| tobiash | good point | 15:48 |
| tobiash | my current prod deployment is based on fddf8f9d | 15:49 |
| corvus | tobiash: while you have it in this configuration, i'd go ahead and set keep to find out what the symlink is, just for later reference. | 15:49 |
| corvus | tobiash: but then after that, i'd suggest upgrading, and if it happens again, let's look at the debug lines from before then to see what roles it previously prepared. | 15:49 |
| tobiash | strange, it prepares an empty dir | 15:51 |
| tobiash | it's not even a symlink | 15:51 |
| corvus | tobiash: oh, hrm, that might be the expected behavior | 15:52 |
| corvus | i think our documentation of the jobdir layout may be misleading | 15:53 |
| corvus | tobiash: is there any chance there is... | 15:54 |
| corvus | tobiash: oh i see it right in the output you sent | 15:54 |
| corvus | tobiash: Prepare zuul role for {'target_name': '' | 15:54 |
| corvus | tobiash: the target name gets appended to the role path root | 15:54 |
| corvus | tobiash: does the job definition say "roles: [{name: '', zuul: '...'}]" ? | 15:55 |
| tobiash | http://paste.openstack.org/show/718949/ | 15:56 |
| tobiash | corvus: that's the job definition | 15:56 |
| tobiash | it works within a different project, but not the project in question | 15:56 |
| corvus | target_name should be 'ansible' there | 15:58 |
| corvus | ah | 15:58 |
| corvus | we don't allow roles named 'ansible' | 15:58 |
| corvus | there's a regex for that in configloader | 15:59 |
| tobiash | the repo is named ansible | 15:59 |
| tobiash | and that's probably what the target_name would be | 15:59 |
| corvus | because we're trying to replace repo names like "ansible_role_foo" with "foo" | 15:59 |
| *** patriciadomin has quit IRC | 16:00 | |
| corvus | tobiash: this is in _makeImplicitRole in configloader | 16:00 |
| *** patriciadomin has joined #zuul | 16:01 | |
| corvus | tobiash: we should probably ensure that leaves us with *something*. if it's empty, maybe just use the pre-regex value. | 16:01 |
| tobiash | corvus: yah | 16:02 |
| tobiash | thanks for finding this :) | 16:02 |
| tobiash | I probably would have searched for hours | 16:02 |
| corvus | sorry i didn't see it right away. this is definitely something that's present in the new code too | 16:03 |
| tobiash | you saw it amazingly fast :) | 16:04 |
| *** harlowja has joined #zuul | 16:05 | |
| openstackgerrit | Tobias Henkel proposed openstack-infra/zuul master: WIP: Fix implicit role for repos named ansible https://review.openstack.org/560554 | 16:12 |
| tobiash | corvus: so I think that should fix it ^ | 16:12 |
| tobiash | (needs test case though) | 16:12 |
| tobiash | will add tests later or tomorrow | 16:16 |
| corvus | tobiash: are you still using https://review.openstack.org/550982 and child? | 16:19 |
| tobiash | corvus: no, that was just to demonstrate an issue in diskimage-builder | 16:26 |
| tobiash | just updated prod to latest master :) | 16:45 |
| *** electrofelix has quit IRC | 16:53 | |
| *** harlowja has quit IRC | 17:07 | |
| Shrews | corvus: So, re: my ZfS changes... move svc mgmt from the Fedora section into the respective nodepool and zuul install sections? | 17:16 |
| *** jpena is now known as jpena|off | 17:16 | |
| Shrews | corvus: and all the directory setup commands for zuul | 17:16 |
| Shrews | what about zookeeper setup? | 17:17 |
| corvus | Shrews: yeah, what do you think about that? i'm thinking that's like all of the boilerplate package installation stuff together -- things you don't really need to think about much, then writing the config can be separate | 17:17 |
| Shrews | corvus: i like it. not sure where the zk stuff fits though | 17:18 |
| corvus | Shrews: maybe do the same for that, and put it first? | 17:18 |
| corvus | Shrews: zk install + setup + svg | 17:18 |
| corvus | er "svc" | 17:19 |
| Shrews | corvus: zk install is repo specific | 17:19 |
| corvus | abbreviation fail | 17:19 |
| Shrews | so i see that in the environment setup part. and that should take care of it's svc file automatically | 17:20 |
| Shrews | so it's really just modifying the config | 17:20 |
| corvus | Shrews: oh, yep. | 17:20 |
| Shrews | which, honestly... why isn't the default config good enough? | 17:20 |
| Shrews | maybe we should just rip that out | 17:21 |
| corvus | Shrews: indeed, good question :) | 17:21 |
| corvus | i haven't checked to see what fedora splats onto disk, it's probably worth a run through | 17:22 |
| *** _ari_ has quit IRC | 17:23 | |
| corvus | i'll check real quick | 17:24 |
| Shrews | oh good, thx. i was about to go through the process of creating a fedora vm | 17:24 |
| corvus | i've got one up, but i need to remove zk first | 17:24 |
| clarkb | zk doesn't have to be repo specific, in fact on suse there is no package so I just install it from upstream's tarballs which includes a sysv init script that should work with systemd | 17:25 |
| corvus | clarkb: yes, but its *so much easier* to install it from distro packages | 17:26 |
| Shrews | oh, my stomach is reminding me I missed lunch. bbiab | 17:28 |
| corvus | Shrews: the answer is there's no default config file on fedora. so at the least, we need to do "sudo cp /etc/zookeeper/zoo_sample.cfg /etc/zookeeper/zoo.cfg". | 17:31 |
| corvus | Shrews: that's really all that needs to happen, so we should probably replace the existing commands with that. | 17:31 |
| *** _ari_ has joined #zuul | 17:32 | |
| clarkb | corvus: do you need to make sure the data dir is created or does the package do that they just don't use the config by default? | 17:32 |
| corvus | clarkb: it does it | 17:32 |
| Shrews | corvus: ack. I can put that after the zk install | 17:32 |
| *** hashar has joined #zuul | 17:38 | |
| tobiash | corvus: I think I have a new nodepool usecase. We like to get github enterprise instances from nodepool for our integration testing. This is supplied as a virtual appliance aka cloud image. So far so good, and easy to do with nodepool. But ghe requires at least one additional volume attached to the instance. Which is I think not yet integrated in nodepool. | 17:44 |
| *** dtruong_ has joined #zuul | 17:45 | |
| tobiash | corvus: what do you think about an 'additional-volumes' config in the pool config? | 17:45 |
| clarkb | tobiash: correct there is no explicit volume management in nodepool today. It can boot instances from volume though (not sure if that solves your problem) | 17:45 |
| *** dtruong_ has quit IRC | 17:45 | |
| clarkb | liek could you boot an instance from volume with enough storage at whatever path ghe expects it at? | 17:46 |
| tobiash | clarkb: nope, I'll have to use that too as the diskimage is larger than my largest flavor | 17:46 |
| tobiash | so I need boot-from-volume and attach another volume | 17:46 |
| *** dtruong_ has joined #zuul | 17:46 | |
| *** dtruong_ has quit IRC | 17:46 | |
| tobiash | github requires a separate disk for user data | 17:47 |
| *** dtruong_ has joined #zuul | 17:47 | |
| *** dtruong_ has quit IRC | 17:47 | |
| tobiash | I think it creates an lvm spanning any non-boot-devices | 17:47 |
| *** dtruong_ has joined #zuul | 17:48 | |
| *** dtruong has quit IRC | 17:48 | |
| *** harlowja has joined #zuul | 17:52 | |
| tobiash | fungi, corvus: commented on 554352 | 18:01 |
| fungi | tobiash: yep, i expect we should try to get at least a third volunteer for better redundancy | 18:03 |
| fungi | though having information on how to securely report suspected vulnerabilities, even if we still need more volunteers to triage them, is probably better than not having those recommendations published anywhere | 18:04 |
| tobiash | fungi: so you like to land it with the todo? | 18:05 |
| tobiash | or just remove the todo? | 18:06 |
| fungi | if there's a better way to leave ourselves a reminder that we want at least one more volunteer there, then i'd be fine removing that todo | 18:07 |
| tobiash | it just feels wrong landing a doc after release with a todo in there | 18:09 |
| corvus | tobiash: i think supporting additional volumes is a good idea. i'm sure Shrews, clarkb, and mordred can help with implementation suggestions. it seems like maybe that should be a pool-label attribute? | 18:14 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: Reorganize "Zuul From Scratch" document https://review.openstack.org/556988 | 18:14 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: Add Gerrit docs to Zuul From Scratch https://review.openstack.org/558600 | 18:14 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: Add static driver doc to Zuul From Scratch https://review.openstack.org/558802 | 18:14 |
| corvus | tobiash, fungi: i plan on adding my key; i may not do it right this moment | 18:14 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: Add sample systemd service files. https://review.openstack.org/558830 | 18:14 |
| openstackgerrit | David Shrewsbury proposed openstack-infra/zuul master: Move zuul jobs config to zuul config section https://review.openstack.org/558867 | 18:14 |
| tobiash | corvus: yes that was the idea | 18:14 |
| Shrews | corvus: end result of that stack looks so much nicer | 18:14 |
| clarkb | tobiash: corvus: yes I think it should be specific to the pool as not all pools will have volumes necessarily | 18:17 |
| tobiash | so we're all on the same page | 18:18 |
| tobiash | let's see when I have time to implement this | 18:18 |
| clarkb | the big thing is like instances and floating IPs these will need to be tracked | 18:18 |
| clarkb | because nova isn't going to reliably delete your volume when nodepool is done with it | 18:19 |
| tobiash | yes and quota checked | 18:19 |
| clarkb | we will either need some way to detect leaks like with floating ips or directly track them as explicit resources like instances | 18:19 |
| tobiash | quota check should even be done today due to boot-from-volume | 18:19 |
| tobiash | clarkb: I think my first try would be to add the volume information to the node in zk and delete it together with the instance | 18:20 |
| clarkb | tobiash: ya like floating IPs | 18:20 |
| tobiash | and maybe add a cleanup thread like the one for leaked instances | 18:20 |
| Shrews | I would like to see all of the various nodepool drivers merged before doing the volume support. | 18:21 |
| clarkb | tobiash: I think you can attach metadata to the volume to indicate it is nodepool managed then if you see one that isn't attached to an instance it can be deleted by leak detection | 18:21 |
| corvus | is it only boot-from-volume that has the option to delete when the instance is deleted? | 18:21 |
| tobiash | corvus: I think so | 18:21 |
| tobiash | but maybe mordred knows better | 18:22 |
| Shrews | iirc, boot-from-volume has a flag to control that (defaults to True) | 18:22 |
| Shrews | i *think* | 18:22 |
| clarkb | I want to say that is a volume level attribute and can eb set for any volume | 18:22 |
| clarkb | so if you attach it to an instance you can say delete this when instance goes away | 18:22 |
| corvus | Shrews: i'm probably going to leave a -2 on the kubernetes/containers drivers for a bit, until we get traction on the container spec. so maybe we can target after the refactor + azure/vmware/ec2 drivers? basically, after all the outstanding vm drivers, i guess? | 18:22 |
| Shrews | corvus: sounds good | 18:23 |
| corvus | clarkb, Shrews, tobiash: so if that delete flag exists, can we rely on it and not have to track the volumes? | 18:23 |
| corvus | "exists for any volume" i should say | 18:24 |
| tobiash | if it does then I think yes as we rely on that already and at least what I see it works without problems so far | 18:24 |
| tobiash | but I haven't seen this option in volume attach personally | 18:24 |
| tobiash | at least the openstack cli doesn't seem to offer this | 18:26 |
| Shrews | it's a nova api maybe? | 18:27 |
| clarkb | corvus: you can rely on it for 99% of the deletes | 18:28 |
| clarkb | and the other 1% will leak because openstack | 18:28 |
| clarkb | then you clean them up like we clean up floating ips | 18:28 |
| clarkb | Shrews: ya I think its nova side /me looks at apis | 18:28 |
| Shrews | block_device_mapping | 18:29 |
| Shrews | block_device_mapping_v2, to be specific about what shade uses | 18:29 |
| tobiash | ah, that sounds right | 18:29 |
| tobiash | overlooked that in the cli | 18:30 |
| Shrews | with a delete_on_termination option in the mapping | 18:30 |
| tobiash | so the additional block device can also be created together with the instance | 18:30 |
| clarkb | ok container spec comments posted looks like Shrews found the api bits first | 18:31 |
| tobiash | and shade seems to support this as well | 18:32 |
| rcarrillocruz | hey folks | 18:32 |
| rcarrillocruz | i've noticed that in zuul | 18:32 |
| rcarrillocruz | on long running shell tasks | 18:32 |
| rcarrillocruz | the console stream logs | 18:32 |
| rcarrillocruz | 'waiting on logger' | 18:32 |
| clarkb | tobiash: Shrews v2 may not support delete_on_termination though? | 18:32 |
| rcarrillocruz | is that normal? | 18:32 |
| rcarrillocruz | any way to prevent that? | 18:32 |
| clarkb | https://docs.openstack.org/nova/latest/user/block-device-mapping.html at least there it isn't documented as supporting it | 18:32 |
| Shrews | clarkb: from past experience developing to those docs, do not believe the docs | 18:33 |
| Shrews | :) | 18:33 |
| clarkb | ha | 18:34 |
| Shrews | they were so frustrating when developing shade. had to just go to the source code most of the time | 18:34 |
| clarkb | shade definitely seems to add it in for v2 so I'm assuming your reading of the source figured that out | 18:35 |
| Shrews | i think mordred actually coded that bit | 18:35 |
| tobiash | should we rather trust shade? | 18:35 |
| tobiash | http://git.openstack.org/cgit/openstack-infra/shade/tree/shade/tests/unit/test_create_server.py#n742 | 18:35 |
| clarkb | tobiash: we should trust it for now but confirm it once we can reasonably test it against a real cloud probably | 18:36 |
| Shrews | corvus: does this match what you were imagining? http://logs.openstack.org/67/558867/2/check/build-sphinx-docs/90556d1/html/admin/zuul-from-scratch.html | 18:41 |
| corvus | Shrews: yeah! | 18:43 |
| clarkb | fwiw I said maybe not worry about getting all the bug fixes in this week (I've got a handful myself queued up) but I'm going to review teh changes tobiash called otu specifically | 18:44 |
| corvus | clarkb, tobiash: yeah, i see those changes as "good to have" but i don't think we should block on them | 18:44 |
| corvus | i think we're about ready -- need to check about the postgres release note | 18:44 |
| clarkb | for me I think as long as we can have a good cadence and openstack acting as canary we can safely release really often | 18:45 |
| clarkb | CI will cover a good chunk of stuff and then production openstack zuul as sanity checker | 18:45 |
| tobiash | corvus, clarkb: regarding the database stuff, 560263 is essential to make the table_prefix work | 18:46 |
| tobiash | shame on me that I found this today in my staging branch :( | 18:46 |
| clarkb | tobiash: oh maybe I start there then | 18:46 |
| tobiash | without that I doubt that anyone can run successfully with table_prefix and more than one instances in one db | 18:47 |
| clarkb | +2 | 18:48 |
| clarkb | corvus: ^ fyi | 18:48 |
| Shrews | rcarrillocruz: i have never seen that, but I don't think that's what we'd expect | 18:49 |
| tobiash | rcarrillocruz: did you start the zuul_stream module in your base job? | 18:50 |
| clarkb | rcarrillocruz: Shrews I want to say what we have seen is a race between the streamer starting in the job and the dashboard likning to it | 18:50 |
| clarkb | seems like tobiash and pabelanger and I debugged it but I was at a conference or something and not sure a patch ever resulted out of that | 18:50 |
| tobiash | clarkb: I just remember you saying 'thing x needs to be moved to y' | 18:51 |
| tobiash | clarkb: however I think thats unrelated to rcarrillocruz's problem | 18:51 |
| clarkb | tobiash: ya sounds like different problem | 18:51 |
| tobiash | rcarrillocruz: do you use the role http://git.openstack.org/cgit/openstack-infra/zuul-jobs/tree/roles/start-zuul-console in your base job? | 18:53 |
| tobiash | if not that might be the problem | 18:53 |
| tobiash | another possible problem could be that your task does things for a long time without starting to print anything | 18:54 |
| tobiash | which could make zuul think there is a problem | 18:54 |
| tobiash | so you might want to add a 'echo foo' at the start to check if that's the case | 18:55 |
| clarkb | ok I've approved the first of tobiash's linked fixes since it was easy. I'm going to make a sandwich for lunch before digging into the others | 18:57 |
| rcarrillocruz | yeah, i think that is used at SF tobiash | 19:04 |
| dmsimard | Should the zuul stream connection timeout eventually ? https://github.com/openstack-infra/zuul/blob/master/zuul/ansible/callback/zuul_stream.py#L125-L126 | 19:06 |
| dmsimard | Saw it try to connect for a good 20 minutes before it gave up | 19:06 |
| rcarrillocruz | for reference: https://ansible.softwarefactory-project.io/logs/11/11/3e2d953bfb56ebb4a31c24bb37e17ada5a6228b6/check/ansible-test-network-integration-vyos-devel/6e1fb5a/job-output.txt.gz | 19:06 |
| rcarrillocruz | look at the 'waitng for logger' lines | 19:06 |
| dmsimard | rcarrillocruz: wait are we asking about the same thing ? lol | 19:07 |
| openstackgerrit | Merged openstack-infra/zuul master: Switch implicit role not found message to debug https://review.openstack.org/556866 | 19:14 |
| tobiash | clarkb: just verified that block device mapping v2 and terminate works | 19:47 |
| tobiash | http://paste.openstack.org/show/718982/ | 19:47 |
| tobiash | it creates a vm using boot-from-volume and adds an additional volume and both get deleted on instance termination :) | 19:48 |
| clarkb | tobiash: perfect so a cleanup cron like for floating IPs is probably sufficient | 19:48 |
| tobiash | yes | 19:48 |
| tobiash | clarkb: however I don't see any information in the docs if we can attach metadata to it | 19:50 |
| tobiash | that maybe needs to be done after creation | 19:50 |
| clarkb | tobiash: or maybe if we can use a name template do that | 19:50 |
| tobiash | also there seems to be no documented way to set a volume name | 19:52 |
| clarkb | :( | 19:54 |
| corvus | clarkb, tobiash: i approved the alembic fix. | 19:58 |
| corvus | clarkb: can you +3 https://review.openstack.org/560079 ? | 19:58 |
| clarkb | corvus: ya just finishing lunch will look | 19:58 |
| corvus | clarkb, tobiash: then i think we should restart openstack-infra after that lands, make sure it doesn't explode, then tag that. | 19:58 |
| tobiash | corvus: sounds good | 19:59 |
| clarkb | corvus: done. And seems reasonable as a plan to me | 20:01 |
| mrhillsman | can someone point me to the code in zuul that talks to this section - https://imgur.com/vPCbL6R - on github and/or the github doc | 20:07 |
| mrhillsman | i say github doc as i think it could be zuul using github3.py, i just do not know what it is called so not sure what to look for | 20:07 |
| mrhillsman | statuses? | 20:11 |
| mrhillsman | yep, nvm | 20:12 |
| openstackgerrit | Merged openstack-infra/zuul master: Also prefix the alembic_version table https://review.openstack.org/560263 | 20:15 |
| *** kmalloc_ has joined #zuul | 20:21 | |
| openstackgerrit | Merged openstack-infra/zuul master: Add postgresql release note https://review.openstack.org/560079 | 20:22 |
| *** dtruong_ has quit IRC | 20:24 | |
| *** dtruong has joined #zuul | 20:25 | |
| *** kmalloc_ has quit IRC | 20:25 | |
| *** kmalloc_ has joined #zuul | 20:25 | |
| *** kmalloc_ is now known as notmorgan | 20:26 | |
| *** notmorgan has quit IRC | 20:26 | |
| *** notmorgan has joined #zuul | 20:26 | |
| *** notmorgan has quit IRC | 20:26 | |
| *** notmorgan has joined #zuul | 20:26 | |
| *** kmalloc has quit IRC | 20:27 | |
| *** andreaf_ has joined #zuul | 20:27 | |
| *** xhku_ has joined #zuul | 20:27 | |
| *** notmorgan is now known as kmalloc | 20:27 | |
| *** mordred has quit IRC | 20:28 | |
| *** andreaf has quit IRC | 20:28 | |
| *** fbouliane has quit IRC | 20:28 | |
| *** andreaf_ is now known as andreaf | 20:29 | |
| *** mordred has joined #zuul | 20:30 | |
| *** acozine1 has quit IRC | 21:01 | |
| openstackgerrit | Merged openstack-infra/zuul master: Update pypi metadata https://review.openstack.org/560098 | 21:26 |
| tobiash | mrhillsman: status checking is not done yet | 21:27 |
| tobiash | mrhillsman: you cal look at https://review.openstack.org/#/c/535680 | 21:28 |
| tobiash | We use that for quite some time now | 21:28 |
| tobiash | It's wip as it has no tests yet | 21:29 |
| clarkb | tobiash: thinking about the volumes problem a bit more. I wonder if we can just use the volume age or modified field (whatever it is) and the fact that it is not attached as our criteria | 21:29 |
| mrhillsman | thx tobiash will have a look, have an interesting use case I am trying to ensure I understand before I bring it up | 21:33 |
| tobiash | clarkb: maybe we also can create the volume beforehand and still use delete on terminate | 21:35 |
| *** hashar has quit IRC | 21:36 | |
| clarkb | tobiash: ya that might be another appraoch and that way we can name it properly or set metadata | 21:36 |
| corvus | tobiash, jlk, sigmavirus24: i just spotted this error from github.com: http://paste.openstack.org/show/718995/ | 21:43 |
| jlk | corvus: I think that was already fixed | 21:50 |
| corvus | jlk: huh... let me check if we failed to upgrade the lib | 21:50 |
| jlk | https://github.com/sigmavirus24/github3.py/pull/824 | 21:50 |
| jlk | it was fixed 9 hours ago, so it missed my release :( | 21:51 |
| corvus | jlk: ah! yeah, we're running 1.1.0 | 21:55 |
| corvus | glad it's fixed! | 21:56 |
| jlk | is it a showstopper for zuul? I could spin up yet another release :D | 21:59 |
| corvus | jlk: that's a good question. based on the pr description, i'd say 'no'. i wonder what the event was though... | 22:02 |
| corvus | it was for https://github.com/ansible/ansible/pull/35920/commits | 22:02 |
| corvus | and it was some kind of a 'pull_request' event, though that's a broad category, and i don't know the action. | 22:02 |
| corvus | (that pull request *does* have a commit -- ie, it's not a plain issue) | 22:03 |
| jlk | yeah, uh... | 22:03 |
| jlk | You have access to the github zuul app, right? within that app page you can find the events that were delivered | 22:03 |
| corvus | oh right! | 22:03 |
| jlk | from there you should be able to see the full event json | 22:03 |
| jlk | also, that MAY not be the same bug, but near the same bug | 22:05 |
| jlk | could have been issue_comment | 22:06 |
| jlk | or issue_label | 22:07 |
| jlk | or it could have been issue_status, but that SHOULD have had a commit associated with it | 22:08 |
| corvus | ok, i think i found the event | 22:08 |
| corvus | action 'labeled' | 22:08 |
| corvus | can i just post this whole json? is that safe? | 22:10 |
| clarkb | corvus: iirc there is a shared secret that zuul confirms to check autenticity of the vent | 22:10 |
| clarkb | *event | 22:10 |
| clarkb | but maybe its hashed I can't remember. That is what I would be concerned about sahring though | 22:10 |
| corvus | i think it's hashed, and i think it's in a header | 22:11 |
| corvus | i think the only info pertaining to us in the data is the installation id, which i assume isn't terribly sensitive | 22:11 |
| corvus | jlk: oh, huh, actually i think this is inside of a further github api call. it's in *response* to the event, but it's not the event itself that's the problem | 22:12 |
| jlk | the secret is in the header | 22:12 |
| jlk | the body should be safe to post | 22:12 |
| corvus | looking at the traceback, this is operating on the result of github.pull_request(...).reviews() | 22:12 |
| jlk | I see, so it's trying to dig through all the reviews. | 22:13 |
| jlk | hrm. | 22:13 |
| jlk | I'm needing to run some errands very soon | 22:17 |
| jlk | BUT | 22:18 |
| jlk | I see where we ASSUME that every pull review has a commit_id, and maybe that's not true. | 22:19 |
| jlk | I don't suppose you can tell WHICH review... | 22:19 |
| corvus | jlk: no, it's in the middle of a list comprehension and doesn't log it. but maybe this is easy to reproduce, assuming no one changes the reviews on that pr in the interim? | 22:24 |
| jlk | yeah it could be easy to test. | 22:25 |
| jlk | I jsut have to remember how to get back to an interactive session to load that PR :D | 22:26 |
| jlk | So what we do is we hit the APi end point for listing the reviews: https://developer.github.com/v3/pulls/reviews/#list-reviews-on-a-pull-request | 22:26 |
| jlk | so /repos/:owner/:repo/pulls/:number/reviews | 22:26 |
| jlk | and we ASSUME that every entry in the list we get will have a 'commit_id' attribute, because we take every entry and build a PullReview object from the data. https://github.com/sigmavirus24/github3.py/blob/develop/github3/pulls.py#L806 | 22:27 |
| jlk | httpie is so useful | 22:30 |
| jlk | https://gist.github.com/omgjlk/0b456ba255e79c6cc675a05926b08f34 | 22:31 |
| -openstackstatus- NOTICE: zuul was restarted to updated to the latest code; you may need to recheck changes uploaded or approvals added between 21:30 and 21:45 | 22:31 | |
| jlk | and here is one without a commit_id https://gist.github.com/omgjlk/0b456ba255e79c6cc675a05926b08f34#file-foo-json-L214-L249 | 22:32 |
| jlk | So it looks like there is SOME situations where a COMMENT pull request review does not have a 'commit_id' associated with it. THANKS GITHUB. | 22:33 |
| jlk | same with an approval. | 22:35 |
| jlk | I'm going to ask internally what the scenarios are. | 22:35 |
| openstackgerrit | Merged openstack-infra/zuul master: Add test for builds/ web route https://review.openstack.org/560117 | 22:41 |
| corvus | jlk: i don't have anything to add, except ++ :) | 22:46 |
| *** tobasco has quit IRC | 22:50 | |
| *** tobasco has joined #zuul | 22:58 | |
| mrhillsman | https://www.irccloud.com/pastebin/3ktRtg5E/ | 23:43 |
| mrhillsman | is this related to the parent: of a job? | 23:44 |
| mrhillsman | trying to understand who the parent and child are here | 23:44 |
| mrhillsman | https://zuul-ci.org/docs/zuul/admin/drivers/zuul.html | 23:45 |
| clarkb | mrhillsman: zuul creates a queues in its pipelines for changes so that they merge serially for example | 23:46 |
| clarkb | mrhillsman: the parent change is the change ahead of the current change in that queue | 23:46 |
| mrhillsman | ok, got it, don't think it will address my use case then :( | 23:48 |
| mrhillsman | back to reading more, thx | 23:48 |
| *** JasonCL has quit IRC | 23:49 | |
| clarkb | mrhillsman: what are you trying to do? | 23:51 |
| mrhillsman | typing it up | 23:53 |
| mrhillsman | i guess it might be thought of as nested pipelines | 23:53 |
| mrhillsman | which we have working | 23:54 |
| mrhillsman | the issue is reporting back to github | 23:54 |
| mrhillsman | let me type example so it is clear | 23:54 |
| mrhillsman | https://hackmd.io/vdbrD8k8QYazKeGzKir8oQ | 23:57 |
| mrhillsman | i hope that makes sense | 23:57 |
| clarkb | mrhillsman: have each pipeline get triggered on the first /test command then configure each pipeline to also trigger on their specific trigger command | 23:58 |
| clarkb | that should be all that is necessary to do that | 23:58 |
| clarkb | I don't know about the reporting side of things | 23:58 |
| clarkb | you can have multiple triggers in each pipeline and any one of them matching results in jobs being run | 23:59 |
| *** JasonCL has joined #zuul | 23:59 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!