Vedran Lerenc

See also https://github.com/gardener/machine-controller-manager/issues/470 for a similar problem on MCM level. We should develop something like a scoring function that tells us whether we have destabilised a landscape or shoot with...

@timebertt That sounds like a great improvement (pushing down the matching into the component that knows best how to do that). I am just wondering, instead of defining the errors...

> But for the sake of progress, I would vote for focusing on the problem at hand first, which includes: removing technical debt, lowering maintenance burden, lowering potential for wrongful...

@rfranzke Sure, I love what @MrBatschner is writing: > TL;DR: The update candidate logic for OS releases whose version number follows semver should be like the logic for Kubernetes versions...

Creation of tens of thousand of clusters is not practical (takes too long, costs too much), so we would have to get creative here. Maybe we could write a "hollow...

Uh, I have severe doubts that there is a truly justified use case (motivational section) and consequently that all this additional complexity is justified. I am sorry. How about the...

@marwinski Is that a possible issue that may lead to the observed, but not necessarily perceived API server outages? They are phantom outages? @istvanballok Wouldn't this be another (good) reason...

That may not only help with: - Our controllers/kubelet get precedence at the shoot API server - Our controllers/gardenlet get precedence at the Gardener API server ...but also with: -...

@pbochynski Yes, we are contemplating the same (cc as well to the already participating Gardener colleagues also @dguendisch, @rfranzke @timebertt ). In general and many years ago we thought it...

Let me add some more details: We need a solution for the following cases: - invalid credentials - deliberate deletion of resources managed by us - creating infrastructure dependencies that...