Cannot save checkpoint with Ray Tune

I create a cluster constructed by two phsical server, a docker container named “ray-head” is run on one server, and a docker container named “ray-worker” is run on the other one. I can run distributed parallel ray tune program. But error occures When I wanna save checkpoint, the details as follows:
ray.tune.error.TuneError: not found after successful sync down. Are you running on a k8s or managed cluster? rsync will not function due to a lack of SSH functionality. You’ll need to use cloud-checkpointing if that’s the case.

But I have no cloud storage, I only own a cluster with two physical servers.
I have spent a whole day on the problem. Please help me out, thanks in advance.

Can you install ssh on both machines? And can you make sure you can ssh from one to the other machine?