Retries and Redriver

Retries

Number of retries for a task can be configured using @Task annotation retries parameter. The default value is 0. Exponential backoff strategy is followed for retries i.e, the wait time between each retry is exponential in seconds e.g 2, 4, 8, 16, 32...... seconds.

All the retries are performed in the same JVM unless there is a JVM or node crash. In case of the crash the remaining retries are done on a different node using Redriver.

Internally retries are performed using Akka scheduler. Refer com.flipkart.flux.impl.task.AkkaTask class for the implementation.

Redriver

Redriver executes stalled/zombie Task instances including those that missed an execution schedule because of node failure in the Flux cluster.