[Spread-users] Handling splits and joins in replication

[Nick Johnson]

I have a simple event server that keeps a list of all events sent to a
particular event queue, and allows clients to ask for ranges. I'm
wanting to implement replication for load-balancing and reliability
using Spread. In order for clients to be able to know that they have
every event, the order of events on each replicated server needs to be
identical. The basic replication using Spread is easy. What I'm unsure
about is how to handle splits/joins.

Has anyone had to deal with this issue? How do you ensure that if a
split occurs, the individual servers won't become out of sync and each
accept events that cannot be merged when a join occurs, without
relying on a single authoritative server (which is hence vulnerable to
failure)?

-Nick Johnson