Vedran Lerenc

@epDHowwD I suggest to get back to your needs, not how you would implement them. It is very difficult for me to follow you, because you make too many assumptions,...

Thank you @epDHowwD. Life update, or in-place update as we call it, doesn't solve the problem. We roll the nodes because that's how we "force" all end-user pods to be...

A problem noticed out-of-band: We cannot let a user statically define requests for control plane pods. This breaks the metering which is requests-based and worse, may create noisy neighbors and...

Some considerations that came up out-of-band: - Shall the user configure that in the spec or via a config map in the cluster (maybe interested parties in this request can...

We collected the following requirements/comments from adopters using/planning to use Gardener on bare metal hardware: - Implement all updates in-place, i.e. preserve the nodes and also the machines (currently not...

Garden Linux PoV: - We will target only Garden Linux for in-place updates. People shall switch to Garden Linux, if they want in-place updates (Garden Linux as differentiator; no time...

@acumino Hi, is that feature now fully implemented? I see the Azure PR is also merged.

/hold FYI: @n-boshnakov Holding until we make https://github.com/gardener/gardener-ai-conformance public.

Yes @pnpavlov ! Thanks.

> We are working on integrating GPU metrics into the Gardener-provided cluster monitoring stack by default in a future release. Line removed @nickytd .