Short answer: this usually means your production environment has an older Prefect version than your local one. In Prefect 3.x,

ProcessPoolTaskRunner

is available at

prefect.task_runners

, so an ImportError typically indicates the prod install predates when this class was included. Quick checks - Please run

prefect version

in both environments and share the outputs. This will confirm the version mismatch. - If you’re using Docker, what base image/tag or requirements pin are you using in prod? Why this happens -

ProcessPoolTaskRunner

is present in modern Prefect (including 3.x). If prod is on an older Prefect 2.x that doesn’t include it, the import will fail there but not locally. - Source showing it exists in current Prefect: prefect/task_runners.py How to fix - Preferred: align the Prefect versions between local and prod (pin the same version). - Upgrade prod to match local (or vice versa), e.g., bump the Prefect version in your requirements/lockfile or update the Docker image tag. - If you must stay on an older Prefect in prod, change your code to use a task runner available in that version (or avoid setting one and use the default). But the most robust approach is to pin a consistent Prefect version across environments. Optional defensive pattern (temporary workaround)

try:
    from prefect.task_runners import ProcessPoolTaskRunner
    task_runner = ProcessPoolTaskRunner(max_workers=4)
except ImportError:
    # Fallback for older installs; consider pinning versions instead of keeping this
    from prefect.task_runners import ThreadPoolTaskRunner
    task_runner = ThreadPoolTaskRunner(max_workers=4)

If it turns out the versions do match and you still see the ImportError, we should check for: - Package shadowing (e.g., a

prefect.py

file or

prefect

directory in your project on prod) - Mixed environments/paths at runtime - How your deployment image is built (sometimes an old layer is cached) Links - Current task runners in source: prefect.task_runners - Prefect 3 release notes: v3.0 release notes - Prefect 2 task runner concepts (for reference if you’re on 2.x): 2.x task runners Send me: - Output of

prefect version

from local and prod - How you deploy (Docker image/requirements) … and I’ll guide you to the exact fix.