Dieter Guendisch
Dieter Guendisch
Sorry for the initial confusion, of course you were right that the error msg I showed was not related (after all, it came from the `NodeUnstageVolume` which is related to...
Another use case is performance tests: relying on a predictable size of the key components would greatly ease them (otherwise perf test users have to extensively warm up the cluster,...
In an internal discussion @oliver-goetz also mentioned the option of allowing customers to specify minimum resource requests while still using VPA, for when the components need more (esp. memory, which...
I guess the only option would be to enable the observability stack for at least a "representative" pipeline cluster and check (while tests run) which of the two suspects (KAPI,...
Does the potential downscaling then apply anytime or only during maintenance windows?
> Yes, but it shouldn't matter for HA clusters anyways, since the PDB prevents evictions of more than 1 member at the same time. I guess that's fine. Yes sure,...
That would be indeed very handy to be able to quickly differentiate in a browser window with multiple tabs with the dashboard (but with e.g. different shoots).
> Wouldn't it be better to have a condition displaying that some MRs are ignored on the shoot itself? This way, the operator will know instantly just by looking at...
I cannot comment on the gains in g/g by reducing the retention to 4 versions, however for our landscapes this would cut away some well needed buffers: - we want...
> I added a minimum lifetime of 14 months Thank you, I'm fine with this now.