Still have some problem with how to make Saga execution coordinator distribued?

[secwang]

Since we have many Open Source distributed log infrastructure, sagalog can be easily distributed and duration. However, how to ensure saga execution coordinator consume sagalog (based sagaid) gently , ensure only one saga chain be dispatch by one saga execution coordinator, and how to fail over when some coordinator go offline. Cause Saga's sub-request may senstive to some system , how to ensure at most once request when saga execution coordinator consume this saga log chain.