Is it somehow possible to force the scheduler to ignore the available resources on a node/cluster while scheduling new pods? We we would like to "overload" our cluster in our lab environment for testing purposes. I could not find anything about it in the docs. Thanks!
This doc may help. You can try to remove the filter PodFitsResources.
There are bunch of feature flags which you can possibly tweak to achieve it but I would say why not use nodeName in the pod spec and effectively bypass the scheduler.
apiVersion: v1
kind: Pod
metadata:
name: nginx
spec:
containers:
- name: nginx
image: nginx
nodeName: kube-01The above pod will run on the node kube-01