how can cluster autoscaler / kubernetes be configured so that autoscaling group ignores blacklisted IPs

[schlichtanders]

Hi there,
I run into the issue that some IPs by Hetzner are blacklisted by some widely used services like container registries. I know that you can use trow to setup a container proxy, however I am afraid that that is not enough for my usecase.

I am building a kubernetes where other clients should be able to run on and hence they may also run into issues with blacklisted IPs. Hence my current optimal goal would be to somehow quickly recognize that a node has blacklisted IPs, so that the cluster autoscaler can quickly get another new machine, hopefully with working IP.

I read a lot of things and it seems what needs to be configured is the binding procedure of the kubernetes scheduler

• either a Filter plugin which identifies the node immediately as buggy
• or a Permit plugin would need to be written which somehow can assess whether the node is valid
• or maybe the default binding algorithm can be configured to quickly recognize that the nodes don't run on a node with blacklisted IP

any help is highly appreciated

[schlichtanders]

After further thoughts I came to the following best approach: #839

it is completely different :) - hence opened as a new idea