I have noticed “ray up” run ray in docker container, --rm docker run options is always set. Is this a good idea that allowing set head node docker container’s restart policy as always, so I won’t need to start head node docker container manually after head node machine restart?
After read Ray 1.x Architecture, it turned out that head node high availabliity is far more complex than simply restart dead head node docker, I’m look forward to known high availability support in future ray relase
Hi Chongshan, we are currently working on Ray head node high availability. I hope we will have experimental support starting next year. I will update this thread when it is ready to be experimented with.