Package com.atlassian.bitbucket.concurrent

Interface BucketedExecutor<T extends Serializable>

Type Parameters:
T - the task type
public interface BucketedExecutor<T extends Serializable>
Executor that groups submitted tasks in 'buckets' and asynchronously processes these buckets of tasks on worker threads. At construction time, a BucketedExecutorSettings is provided that defines the maximum concurrency level, the task bucketing logic and the BucketProcessor that's called for processing buckets of tasks.

  • Method Summary

    Modifier and Type
    Method
    Description
    void
    schedule(T task, long delay, TimeUnit timeUnit)
    Add a task to be processed and schedule the BucketProcessor to process the task's bucket with a delay.
    void
    shutdown()
    Shuts down the executor and releases all resources acquired by the executor.
    void
    submit(T task)
    Submit a task for processing and schedule the BucketProcessor to process the task's bucket immediately.

  • Method Details

    • schedule

      void schedule(@Nonnull T task, long delay, @Nonnull TimeUnit timeUnit)
      Add a task to be processed and schedule the BucketProcessor to process the task's bucket with a delay. The task may still be processed earlier if a prior task for the same bucket has been scheduled to run before the scheduled time.
      Parameters:
      task - the task to submit
      delay - the processing delay
      timeUnit - the delay's timeUnit

    • shutdown

      void shutdown()
      Shuts down the executor and releases all resources acquired by the executor. This shuts down the executor on the current node only. In a cluster, the corresponding executors will keep processing tasks until they're shut down.

      Note: All plugins which request a bucketed executor must shut it down when the plugin is disabled. Failing to do so will leave behind orphaned instances. On plugin upgrade this can be particularly problematic given the fact that when the new version of the plugin requests an executor it will be provided with the old stale instance which may trigger unexpected behavior, like ClassCastExceptions caused by ClassLoader mismatches. This behaviour will continue until the next time the system restarts (in a clustered environment this means every node).

    • submit

      void submit(@Nonnull T task)
      Submit a task for processing and schedule the BucketProcessor to process the task's bucket immediately. This is equivalent to calling schedule(Serializable, long, TimeUnit) with a delay of 0.
      Parameters:
      task - the task to submit