to force a max number of pods with a given label per node, and can't find anything relevant. Maybe someone here did something like this before? I really don't want to get into writing my own custom scheduler just for this...
Pod Topology spread might help, haven’t tried personally https://kubernetes.io/blog/2020/05/introducing-podtopologyspread/
nothing about a per-label count in there unfortunately
^ (Did not try it myself but it’s something I though of a while ago on similar issue)
oh this looks really cool! Do you know whether cluster autoscaler will get triggered by this?
I think it is limited to cpu and mem the last i’ve checked But indeed a valid point to test
thanks, this gives me a reason to stop pulling out hair for a while :)
Np :) Do update if it makes the cut
some initial tests look really good. The one concern is having to add the parameter to each node manually via curl, need to find a way to automate this, or autoscaler will go moot
Обсуждают сегодня