Sync from the discovery engine
Treydone opened this issue · comments
Most of discovery engines (as Zookeeper as the default one in case of distributed mode) own now their own keyvalue store. As for Apache Storm, the sync should be done from the discovery engine. A controller tries to acquire a distributed look, update the various keys in the distributed engine, then the workers and the frontiers update their configuration from it. Less RPC calls and a better scalability.