Currently, Flyte does not support re-running only the aborted subtasks within a map task directly. When a map task is retried, it typically retries the entire map task, including all subtasks, rather than just the failed ones. This behavior is due to the way Flyte handles retries at the task level, rather than at the subtask level within map tasks. According to
Flyte's documentation, map tasks execute many instances of a task within a single workflow node, and retries are generally managed at the task level.
Would you like more detail on how to handle retries or manage task failures in Flyte?
Sources:
-
Flyte's documentation 👍 0 button 👎 0 button 💡
Hint: Mention @runllm in the thread for followups.