These are chat archives for atomix/atomix

2nd
Apr 2017
Jordan Halterman
@kuujo
Apr 02 2017 23:10
Sure it would serve another potential use case, but it's far to big a refactoring to take on for such little gain. It's not impossible to remove a node from a cluster right now. Supporting a cluster that's not configurable would require a major refactoring of the interfaces (which expose configuration methods) and the internals of the Raft implementation, and that introduces potential instability. If it's feasible at all, it would have to be done in the next major version. And until then, all that can be done is allowing a configuration to be overridden at startup.
Jordan Halterman
@kuujo
Apr 02 2017 23:16
We have a lot more important fish to fry. If you want to take it on and can logically separate configuration changes from the Raft implementation and it passes our QA, I'll accept it. But I wouldn't recommend it. Much higher on my list of priorities is rewriting the log, rewriting the transport, removing serialization, improving the threading model, supporting purely snapshot based state compaction, etc. Perhaps Copycat 2.0 can have an architecture that can better support different approaches to cluster management.
I’ll be in the bay area for a confernce this week but I should still be around to chat some more