This is my design of architecture as shown in figure, in theory, any machine downtime, cluster can still work normally, don't know whether this is feasible,
The great spirit guide,

CodePudding user response:

This is a figure, forgot to post...