Vedran Lerenc
Vedran Lerenc
> So, it appears to me as if maximum is better left regional and only maxSurge and maxUnavailable make much more sense to be looked at in a zonal way....
@oliver-goetz Yes, this has disadvantages. Another that @unmarshall brought up is that a zone may be out of resources - that would also block things. Just saying, you are both...
> Isn't that something we could optimize, .e.g. the error would be detected, the rollout in that zone paused and the next zone is picked for rolling update? Just a...
👍 to @vpnachev 's proposal to avoid replication into the seed and run `SubjectAccessReview` directly against the virtual garden cluster, which must be always reachable from a seed.
I love it ❤️. Great 🚀! @HappyTobi In comparison to "locking", we worried people will automate it, so 4-👀 "dual-approval" seemed more safe. As this happened to you or your...
Shoot resource status.lastOperation.description can get too large (easily reach resource size limit)
I added the label `enhancement` rather than `bug`, but this behaviour may feel to the affected people more like a bug, because they a.) rarely know what the error messages...
> we should check the the error code of Driver.GetMachineStatus call to see if the VM instance has gone. The problem with that is that infrastructures such as Azure are...
FYI: Proposal https://gist.github.tools.sap/D043832/f5d0ac0e0bb138eea07c84ac79e10ce9#step-1-enhancing-zone-distribution-and-surge-management (only step 1 relevant here)
Good question, but I guess on create is sufficient and simpler to build (with an admission plugin to prevent changing the tags in the shoot spec)? It would be a...
OK, then let's focus on immutable tags. That we then initiate a rolling update is the right thing. 👍