Zero nodes available for deployment with 30 nodes while not fully used

Hi @enzoferey , we’re about rolling Karpenter to your cluster once it’s stable (cf this thread) - this kind of issue will no longer be a problem. It’s due to the default Kubernete nodes autoscaler that does not properly schedule pods based on the real workload.

If you want to better understand what’s the issue you’re currently facing, I invite you to read this detailed thread.