Worker nodes should be safer by default, conform to the CIS Kubernetes Benchmark #427
Labels
kind/cleanup
Categorizes issue or PR as related to cleaning up code, process, or technical debt.
sig/cluster-management
Denotes a PR or issue as being assigned to SIG Cluster Management.
I first wrote this in the kubermatic project put then realized this should probably go here so I copied the text. While my primary concern when writing this was with a KKP UC's worker nodes, and the output below is from a Kubernetes 1.31.3 cluster on KKP 2.26.1, I have the same concerns about KubeOne cluster worker nodes.
Description of the feature you would like to add
Once again I've run the CIS Kubernetes Benchmark (specifically Aqua Security's
kube-bench
tool, executed through the job manifest) and while I think you've addressed some open points in the past, several are still open. It's rather concerning particularly because they seem to be low-hanging fruits, as far as my understanding goes, but maybe things are more complicated than they seem.Our end users are getting louder about this as well, as the focus on security in general keeps increasing for good reason.
Solution details
Nothing specific, but all the FAILs of the kube-bench run should be addressed (bonus points for any WARN that is addressed as well, if there's anything to do):
Note that the benchmark also another suite of tests aimed at policies rather than nodes, which I consider out of scope here (and didn't include the output above) - this is just about the node security.
Alternative approaches
Maybe we could solve most of these ourselves with a custom OSP, but not sure that's the right place for everything in the first place.
Use cases
Well, security is important, and the CIS Kubernetes Benchmark is well recognized.
Additional information
The text was updated successfully, but these errors were encountered: