Short answer: if you’re using Prefect Cloud’s Job Run v2 on Prefect Serverless, no — the compute is ephemeral and we don’t expose a toggle to keep the container/pod around after a crash. Logs and run metadata are preserved in Prefect Cloud for debugging. If you need to preserve the underlying job for inspection, run on your own worker instead of Serverless and adjust the infrastructure settings: - Kubernetes worker: edit your work pool’s base job template to prevent automatic cleanup. - Omit

spec.ttlSecondsAfterFinished

(or set a large value) so Jobs aren’t deleted immediately - Optionally set

spec.backoffLimit: 0

to avoid retries and keep the first failed pod intact - Example base job template snippet:

apiVersion: batch/v1
    kind: Job
    spec:
      backoffLimit: 0
      # omit ttlSecondsAfterFinished to avoid auto-cleanup, or set a large value
      template:
        spec:
          restartPolicy: Never
          containers:
            - name: prefect-job
              # image/command/env are injected by the worker

- Configure this in Cloud: Work Pools -> your Kubernetes pool -> Base job template. - Docs: Kubernetes worker guide and Managing work pools - Docker worker: you can configure the base container settings to avoid auto-removal so failed containers stick around for inspection. If you’re on Docker, I can share an exact template. Regardless of compute, you can pull full logs for a failed run:

prefect flow-run logs <flow-run-id>

CLI ref: prefect flow-run logs Can you confirm which compute you’re using for Job Run v2 — Prefect Serverless, a Kubernetes worker, or Docker? I’ll tailor the exact settings/template for your setup.