== [[FetchFailedException]] FetchFailedException

FetchFailedException exception executor:TaskRunner.md#run-FetchFailedException[may be thrown when a task runs] (and storage:ShuffleBlockFetcherIterator.md#throwFetchFailedException[ShuffleBlockFetcherIterator did not manage to fetch shuffle blocks]).

FetchFailedException contains the following:

  • the unique identifier for a storage:BlockManager.md[BlockManager] (as storage:BlockManagerId.md[])
  • shuffleId
  • mapId
  • reduceId
  • A short exception message
  • cause - the root Throwable object

When FetchFailedException is reported, executor:TaskRunner.md#run-FetchFailedException[TaskRunner catches it and notifies ExecutorBackend] (with TaskState.FAILED task state).

The root cause of the FetchFailedException is usually because the executor:Executor.md[] (with the storage:BlockManager.md[BlockManager] for the shuffle blocks) is lost (i.e. no longer available) due to:

  1. OutOfMemoryError could be thrown (aka OOMed) or some other unhandled exception.
  2. The cluster manager that manages the workers with the executors of your Spark application, e.g. YARN, enforces the container memory limits and eventually decided to kill the executor due to excessive memory usage.

You should review the logs of the Spark application using webui:index.md[web UI], spark-history-server:index.md[Spark History Server] or cluster-specific tools like https://hadoop.apache.org/docs/stable/hadoop-yarn/hadoop-yarn-site/YarnCommands.html#logs[yarn logs -applicationId] for Hadoop YARN.

A solution is usually to tune the memory of your Spark application.

CAUTION: FIXME Image with the call to ExecutorBackend.

=== [[toTaskFailedReason]] toTaskFailedReason Method


Back to top