>>> We sleep long enough when gossiping pending ranges before starting to >>> move data that we're safe from micropartitions. >> >> Hmm... I don't think we sleep at all at that time. Once we get load >> info, we gossip pending ranges and start to move data immediately. > > You're right. That's a bug IMO.
This makes things significantly clearer to me :-) This is exactly the reason why I was reluctant to release beta, as I felt pending ranges & simultaneous bootstrap are too unreliable. I created #575 and submitted a patch (sleep for RING_DELAY between gossip & bootstrap). With this included, I think most of the things I had doubts about in bootstrap will be gone.
