| -@gerrit:opendev.org- Albin Vass proposed: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 09:50 | |
| -@gerrit:opendev.org- Albin Vass proposed: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 09:57 | |
| -@gerrit:opendev.org- Zuul merged on behalf of Benjamin Schanzel: [zuul/zuul] 911700: Web UI: substring search for builds, buildsets https://review.opendev.org/c/zuul/zuul/+/911700 | 10:27 | |
| -@gerrit:opendev.org- Zuul merged on behalf of Simon Westphahl: [zuul/zuul] 913875: Correctly limit buildsets with multiple refs https://review.opendev.org/c/zuul/zuul/+/913875 | 10:48 | |
| -@gerrit:opendev.org- Simon Westphahl proposed: [zuul/zuul] 914512: Log Zookeeper session on connection state change https://review.opendev.org/c/zuul/zuul/+/914512 | 11:20 | |
| -@gerrit:opendev.org- Albin Vass proposed: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 11:29 | |
| -@gerrit:opendev.org- Zuul merged on behalf of Simon Westphahl: [zuul/zuul] 914512: Log Zookeeper session on connection state change https://review.opendev.org/c/zuul/zuul/+/914512 | 15:04 | |
| @clarkb:matrix.org | Albin Vass: re 914434 I wonder if there are some node types we should be adding too? Having a safe fallback like your change also makes sense but the problem implies there are types we don't currently handle which we maybe should? | 15:05 |
|---|---|---|
| @sylvass:albinvass.se | The nodes that were causing the failures were not on-demand or spot instances, but dedicated instances (Is4gen in AWS). So I don't think it makes sense to add them | 15:11 |
| @sylvass:albinvass.se | Or at least I couldn't find a quota code for on-demand or spot for those | 15:13 |
| @clarkb:matrix.org | I don't know enough about amazon quotas to say, but my inclination would be to handle explicit cases like that explicitly even if we then indicate they don't count towards quotas | 15:14 |
| @clarkb:matrix.org | and rely on the fallback case only for truly unexpected cases | 15:14 |
| @jim:acmegating.com | it seems there are 3 types of instances -- on-demand, spot, and dedicated hosts; we only support on-demand and spot, and there's a whole bunch of quota codes for the dedicated hosts instance types. it would be good to add them, but we'd probably need to add support for dedicated host types in order to actually include that in quota calculations. i think the proposed fix is a good first step, and we can leave expanding the quota codes to a future change since it's likely to be rather involved. | 15:20 |
| one thing to note is that the original behavior was to fall back without error, but i believe the addition of spot instances broke the fallback behavior, so this change will just restore what we had before. however, i do have some changes i'd like to make to Albin Vass 's change that i should be able to get to shortly. | ||
| @jim:acmegating.com | i'm wondering if `Running Dedicated i4i Hosts` code `L-0300530D` is the right one for IS4gen? i'm not seeing anything else involving "I" and "4" that seems likely... | 15:22 |
| @jim:acmegating.com | (it looks like there are 53 dedicated host quota codes) | 15:23 |
| @clarkb:matrix.org | I guess quota codes are the canonical way to lookup a quota cost? And to properly use them we have to map a node type to the quota code and for some reason this is more straightforward for spot and on demand instances? | 15:30 |
| @jim:acmegating.com | Clark: yes they are canonical, and yes we have to map the instance type see https://opendev.org/zuul/nodepool/src/branch/master/nodepool/driver/aws/adapter.py#L53 | 15:39 |
| @jim:acmegating.com | Clark: it's not more straightforward for the others, it's just that to actually use it in nodepool, we would need to add support for dedicated hosts and add a third column there, and manually identify which quota codes go with which instance families, which, given the ambiguity in the example above ("i4i" hosts) might actually entail spinning up a bunch of dedicated hosts to confirm our guesses. that might entail actual monetary expenditure. i don't think we need to do that unless/until someone actually wants to spin up dedicated hosts with nodepool. | 15:42 |
| @clarkb:matrix.org | gotcha | 15:45 |
| @jim:acmegating.com | (we could probably go ahead and add the instance type for `is` (`is4gen` would go through our regex as `is`) to that list with empty values for both columns (like hpc has an empty string for spot) if we are sure of the instance type) | 15:48 |
| @jim:acmegating.com | * (we could probably go ahead and add the instance type for `is` (`is4gen` would go through our regex as `is`) to that list with empty values for both columns (like hpc has an empty string for spot) if we are sure of the instance type) (but i'm not 100% sure on that) | 15:48 |
| @harbott.osism.tech:regio.chat | is it intended that using multiple filters for config-errors doesn't work? see e.g. https://zuul.opendev.org/t/openstack/config-errors?branch=unmaintained%2Fvictoria&branch=unmaintained%2Fwallaby&skip=0 | 15:52 |
| @harbott.osism.tech:regio.chat | I would like to have a single URL to show all issues with unmaintained branches | 15:53 |
| @clarkb:matrix.org | I think those filters may be ANDed together not ORed. Allows you to be more specific and drill down but not capture a variety of errors in one go | 15:54 |
| @harbott.osism.tech:regio.chat | hmm. it does work for other searches like builds https://zuul.opendev.org/t/openstack/builds?branch=unmaintained%2Fvictoria&branch=unmaintained%2Fyoga&skip=0 | 15:55 |
| @clarkb:matrix.org | Builds do a SQL query but errors are an internal filtering. That might explain the difference? | 15:57 |
| @jim:acmegating.com | i dug a little more, and i think is4gen is available as an on-demand instance, but i can't find a quota code for it :/ | 16:09 |
| @jim:acmegating.com | okay i launched one, and i don't see any quota usage for it | 16:19 |
| @jim:acmegating.com | so for the moment, i don't think there's anything we could add to that table if we wanted to :/ | 16:19 |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed on behalf of Albin Vass: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 16:26 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed on behalf of Albin Vass: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 16:27 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed on behalf of Albin Vass: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 16:39 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed on behalf of Albin Vass: [zuul/nodepool] 914434: Fix issue with unknown instance family https://review.opendev.org/c/zuul/nodepool/+/914434 | 18:51 | |
| @sylvass:albinvass.se | Clark: re https://review.opendev.org/c/zuul/zuul-client/+/913797 | 19:11 |
| is it possible to run pbr as a cli to get the current version? If so it should be easy to sync the versions | ||
| @jim:acmegating.com | Albin Vass: `pbr info zuulclient` in an installed env i think | 19:14 |
| @sylvass:albinvass.se | cool, i'll try that | 19:15 |
| @jim:acmegating.com | Albin Vass: i think it might be `pbr info zuul-client` | 19:16 |
| @jim:acmegating.com | the pypi and python package names differ in that one | 19:16 |
| @sylvass:albinvass.se | oh interesting, you have to build the package to get the package version | 19:39 |
| @sylvass:albinvass.se | you can get it from `python3 setup.py --version` as well | 19:42 |
| @clarkb:matrix.org | I think setup.py is being deprecated for stuff like that but not a problem yet | 20:10 |
| @sylvass:albinvass.se | okay this got annoying. nix flakes don't copy the .git directory over to the store (because it's not binary reproducible), so it's not possible to get the version from git. | 20:17 |
| It is possible to use the commit hash with `self.rev` but that's not helpful in this case. | ||
| Technically in nix it doesn't matter much because the only thing that is affected is `zuul-client --version` outputs the wrong version. | ||
| @sylvass:albinvass.se | you reference flakes through git urls anyway, and if you want a specific tag you specify that in the url. | 20:18 |
| -@gerrit:opendev.org- Albin Vass proposed wip: [zuul/zuul-client] 913797: Add nix flake https://review.opendev.org/c/zuul/zuul-client/+/913797 | 20:21 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: [zuul/zuul] 914415: Move fake gerrit and pagure into dedicated files https://review.opendev.org/c/zuul/zuul/+/914415 | 21:27 | |
| -@gerrit:opendev.org- James E. Blair https://matrix.to/#/@jim:acmegating.com proposed: | 23:19 | |
| - [zuul/zuul] 914561: Make the test change database serializable https://review.opendev.org/c/zuul/zuul/+/914561 | ||
| - [zuul/zuul] 914562: WIP Add an upgrade test https://review.opendev.org/c/zuul/zuul/+/914562 | ||
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!