#3925 [Core feature] maptask executions should be recoverable Issue created by hamersaw Motivation: Why do you think this is important? Currently

maptask

executions are completely recomputed in the event of a workflow recovery. This is because subtask execution state is maintained using an

ExternalResourceInfo

field within the

TaskExecutionClosure

. The recovery mechanism only looks at the top-level status, namely whether the entire task has completed successfully or not, and therefore cannot recover individual successful subtask executions. Goal: What should the final outcome look like, ideally? The workflow recovery paradigm should work with

maptasks

. Since the introduction of ArrayNode this is possible because recovery can be performed at the node-level. An

ArrayNode

could query state of each subtask and initialize the subtask execution metadata by recovering previously successful subtask executions. Describe alternatives you've considered Leaving it as is may result in unintended behavior and unnecessary duplication of subtask executions on workflow execution recovery. Propose: Link/Inline OR Additional context No response Are you sure this issue hasn't been raised already? ☑︎ Yes Have you read the Code of Conduct? ☑︎ Yes flyteorg/flyte