medium10 pts · first answer countsml-pipelinesorchestration

Safe retries in pipelines

An orchestrator retries a failed feature pipeline task, and the result table now has duplicate rows for that day. What design principle was missed?

A.Tasks should never be retriedB.Each task should run on a new clusterC.All tasks should write to the same tableD.Tasks should be idempotent, for example overwriting the partition for that date instead of appending

Discussion 0

Join free to post in the discussion.

Safe retries in pipelines

Discussion 0

Related questions