| -@gerrit:opendev.org- likui proposed: [zuul/nodepool] 804837: setup.cfg: Replace dashes with underscores https://review.opendev.org/c/zuul/nodepool/+/804837 | 07:39 | |
| @dpawlik:matrix.org | hello folks o/ Quick question: is possible to remove the image UUID from the nodepool launcher cache? From time to time I need to remove update the cloud-image in Openstack, but the nodepool-launcher keeps the UUID in cache for old one, so I need to remove old one to spawn instance with new image. I would like to avoid situation, that I'm removing old image, so my question is: is possible to clean or change the image UUID in nodepool launcher cache? | 07:57 |
|---|---|---|
| @dpawlik:matrix.org | corvus: hey, that change https://review.opendev.org/c/zuul/zuul-jobs/+/894755 will be helpful to merge https://review.opendev.org/c/zuul/zuul-operator/+/881245 | 12:20 |
| @fungicide:matrix.org | > <@dpawlik:matrix.org> hello folks o/ Quick question: is possible to remove the image UUID from the nodepool launcher cache? From time to time I need to remove update the cloud-image in Openstack, but the nodepool-launcher keeps the UUID in cache for old one, so I need to remove old one to spawn instance with new image. I would like to avoid situation, that I'm removing old image, so my question is: is possible to clean or change the image UUID in nodepool launcher cache? | 12:44 |
| are you using a nodepool-builder to build those images? i don't recall ever having to touch any nodepool-launcher cache in opendev's deployment. the launchers just always use the newest uploaded image for a given label, and the builders clean up old images in excess of the configured retention | ||
| @clarkb:matrix.org | > <@fungicide:matrix.org> are you using a nodepool-builder to build those images? i don't recall ever having to touch any nodepool-launcher cache in opendev's deployment. the launchers just always use the newest uploaded image for a given label, and the builders clean up old images in excess of the configured retention | 15:11 |
| My interpretation of the question is that dpawlik is using a cloud image that isn't managed by nodepool. The image is used by name and the images are replaced. I think there are two options here. The first is just wait for the cache to clear (I want to say it is a 5 minute cache by default) and have failed boots until then. Or update the nodepool config to put a new image in place instead and then clean up the old image rather than the other way around. This is what I would do | ||
| @clarkb:matrix.org | Does anyone know what controls whether or not a reply in matrix starts a thread or is just a quoted reply in the main channel? It seems like when I do it it is a quoted reply. Which I actually prefer. The reason I prefer it is that matrix threads are super understated and makes it really hard to determine where unread messages are | 15:45 |
| @clarkb:matrix.org | I feel like it has a lot of the same problems slack threads have... | 15:45 |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of Ian Wienand: [zuul/zuul] 794688: github: fallback to api_token when can't find installation https://review.opendev.org/c/zuul/zuul/+/794688 | 16:23 | |
| @clarkb:matrix.org | that is just a rebase to address the merge confligt. Going to look at addressing comments next | 16:24 |
| @fungicide:matrix.org | > <@clarkb:matrix.org> Does anyone know what controls whether or not a reply in matrix starts a thread or is just a quoted reply in the main channel? It seems like when I do it it is a quoted reply. Which I actually prefer. The reason I prefer it is that matrix threads are super understated and makes it really hard to determine where unread messages are | 16:25 |
| in the element client there are two adjacent context buttons when you hover over a message: "reply" and "reply in thread" | ||
| @clarkb:matrix.org | oh heh I had always just used the reply button and didn't realize the one next to it offered the thread reply | 16:28 |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of Ian Wienand: [zuul/zuul] 794688: github: fallback to api_token when can't find installation https://review.opendev.org/c/zuul/zuul/+/794688 | 16:45 | |
| @clarkb:matrix.org | corvus: ianw ^ how does that look? There was a fair bit of discussion and there were also the review comments. I tried to synthesize things into this patchset but please let me know if I misparsed or missed anything | 16:45 |
| -@gerrit:opendev.org- Clark Boylan proposed on behalf of Ian Wienand: [zuul/zuul] 794688: github: fallback to api_token when can't find installation https://review.opendev.org/c/zuul/zuul/+/794688 | 17:05 | |
| @jim:acmegating.com | Clark: lgtm | 17:20 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 892416: Make Ansible variable freezing more efficient https://review.opendev.org/c/zuul/zuul/+/892416 | 20:32 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 892416: Make Ansible variable freezing more efficient https://review.opendev.org/c/zuul/zuul/+/892416 | 20:43 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 892416: Make Ansible variable freezing more efficient https://review.opendev.org/c/zuul/zuul/+/892416 | 21:00 | |
| @iwienand:matrix.org | i am super super hesitant to -1 but i put a longish comment in. I only did because i am sure i am wrong about it, but in my wrongness i think probably it's worth updating the CL or documentation to explain why i'm wrong :) | 22:05 |
| @clarkb:matrix.org | I think what I'm confused about is we do have a number of repos without app installs that seem to be working at least to some extent | 22:23 |
| @clarkb:matrix.org | And ya maybe depends on or some other features do need a token but what we are doing doesn't appear to. Therefore making it required seems overkill | 22:24 |
| @clarkb:matrix.org | I guess what we are doing typically doesn't involve changes/PRs. It is only keeping a local cache of the GitHub repo state and providing it to jobs | 22:26 |
| @clarkb:matrix.org | And that avoids the whole update change path that needs graphql requests | 22:26 |
| @jim:acmegating.com | okay, i see the code path that can trigger that in check now. | 22:31 |
| @clarkb:matrix.org | Ya I think it does exist. But I'm not convinced that is a reason to make the token required. More strongly suggested? | 22:32 |
| @jim:acmegating.com | ianw: so switching the docs back to "should create" in order to support depends-on would resolve the issue, right? | 22:32 |
| @jim:acmegating.com | i agree, i think maybe the code is good, and the docs / release note should just clarify that for depends-on support (ie, anything with PRs) you either need an installation or a token | 22:33 |
| @jim:acmegating.com | (the code could probably be further improved to return some kind of error if zuul goes down that code path without auth, but that seems going above and beyond for this change) | 22:34 |
| @iwienand:matrix.org | thinking through why we have this in system-config, is that we have a devel job that installs ansible/ara and we want to install from the zuul git checkout of main/master which zuul will grab from github. so that is ok unauthenticated. but the caveat is that without an auth token, we can't depend-on changes from those projects. so perhaps it should be explained in practical (but more generic) terms like that? | 23:15 |
| @jim:acmegating.com | yeah i said something similar in a reply to your comment | 23:19 |
| @iwienand:matrix.org | ah yes, indeed ... that split up makes sense | 23:20 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 894839: Use a seeded PRNG when setting timer triggers https://review.opendev.org/c/zuul/zuul/+/894839 | 23:41 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!