Default tolerations are not getting applied to single user pods


In order to make cluster autoscaler work efficiently with Jupyterhub I am following this documentation about optimizations.

From which I found this doc which mentions about the default tolerations available for a single user notebook pod. These tolerations are present in image-puller and user-placeholder pods but somehow single user notebook pod don’t get these tolerations.

When I try to create a notebook, I get 1 node(s) had taints that the pod didn't tolerate, 2 node(s) didn't match node selector.

I have two nodes which are tainted one of them is GPU and other one is CPU. Tainted CPU node should be able to schedule single user pods.

I have written a custom kubespawner and I was overriding the default tolerations :sweat_smile: that is why they were never propagated properly. Thanks :slight_smile: