Class ShardedDaemonProcessCoordinator

  • public final class ShardedDaemonProcessCoordinator
    extends java.lang.Object
    Rescaling workflow overview:

    1. stop pinging workers 2. Write to ddata new size and revision, in progress/not complete 3. Stop shards for all workers via shard coordinator 4. Update ddata state, mark rescaling completed 5. start pinging workers again

    If coordinator moved before completing it will re-trigger workflow from step 2 on start, worst case stopping all workers before starting them again.