Java client for the NSQ realtime distributed messaging platform
MIT License
73
stars
25
forks
source link
DP-12: When all connections get stuck in a FAILED state, clear / reset them all so we don't intentionally drop subsequent publishing calls on the floor. #52
Before our Publisher refactorings, the old client behavior used to ping / pong back and forth between the primary and failover, so that we'd never get stuck in a state where we weren't at least attempting to reconnect on subsequent publishing attempts.
This attempts to approximate that behavior by resetting all the connection states back to NOT_CONNECTED when everything is marked as FAILED.
Before our Publisher refactorings, the old client behavior used to ping / pong back and forth between the primary and failover, so that we'd never get stuck in a state where we weren't at least attempting to reconnect on subsequent publishing attempts.
This attempts to approximate that behavior by resetting all the connection states back to NOT_CONNECTED when everything is marked as FAILED.