We have been running into an issue with the managed kubernetes cluster not deleting old docker images and filling up the disk space on the nodes causing new deployments to fail when pulling down new images casing an “No space left on device” error.

Reading into the kubernetes documentation the garbage collection should take care of deleting the unused images but this doesnt appear to be the case.

Is there a way to confirm the GC is running on each node and if it is not enable it?

Our current solution to resolve the problem is to setup a new pod that connect the the nodes docker instance and run the delete images command manually.