Caleb Bron

@jiaqiluo some design comments: > Along with a new action named setpodsecurity on the object: We should avoid using action's, that was a design pattern influenced by the cluster manager...

@Shavindra thats a project, not a namespace. Also if you made the user a project-owner of the project, they should be able to update the project they own. This line:...

The only way I can think of this working is to add a cache in between steve and etcd, and that sounds like a nightmare to me. Caching bugs would...

Can you add details on which items you are requesting and how ? My guess it that Norman is scoping/filtering more by default (ie like a project level) and so...

Thanks, that gives us something to go off of. Are all your deployments then in just 1 project ? Cause the second call isn't filtering by project. Under the API...

An external kubectl to the cluster still works correct, it's only kubectl from within the shell that is broken ? And I'm assuming if coredns is enabled it still works....

Either @steffantucker or @MbolotSuse can pick up, it has clear path forward so Tucker just ask michael about it first.

`Creating 1000 cluster roles with circular reference with older version of the webhook` This is interesting test. That being said, I don't think it could happen in the real world....

Team 1 should be fine here, as there are no k8s deletions in 1.24, and both charts are not capped until 1.25: * csp-adapter: https://github.com/rancher/charts/blob/dev-v2.6/charts/rancher-csp-adapter/0.1.0%2Bup0.1.0-rc8/Chart.yaml#L5 * webhook: https://github.com/rancher/charts/blob/dev-v2.6/charts/rancher-webhook/1.0.5%2Bup0.2.6-rc3/Chart.yaml

Ideally this will make 2.6.7, if not our tests wouldn't be legitimate.