Open drewrobb opened 9 years ago
Sorry for late response. Agreed, yes nimbus is not ready for high availability mode, my initial thought was that Nimbus is stateless. My two cents: Just wondering if it is possible to maintain state of Nimbus on Zookeeper/Redis etc, or maintaining the state on a separate Docker container (hosted). I think this is a good challenge to resolve !
Nimbus needs to store submitted topologies locally in storm.local.dir. Failover in marathon will start with an empty directory and topologies will not be able to start new workers. There is no high availability mode for nimbus yet. See https://issues.apache.org/jira/browse/STORM-166