Yes — in Prefect 3.x you can retry a flow run by ID through the API by setting its state again. Python API example:

python
from uuid import UUID

from prefect.client.orchestration import get_client
from prefect.states import Scheduled
from prefect.client.schemas.responses import SetStateStatus

async def retry_flow_run(flow_run_id: str):
    async with get_client() as client:
        result = await client.set_flow_run_state(
            flow_run_id=UUID(flow_run_id),
            state=Scheduled(message="Retrying via API"),
            force=True,
        )

        if result.status == SetStateStatus.ACCEPT:
            return result.state
        else:
            raise RuntimeError(f"Retry was not accepted: {result.details}")

What this does -

set_flow_run_state(...)

changes the run’s state - using

Scheduled(...)

puts it back in the queue -

force=True

bypasses orchestration checks that might otherwise reject the transition REST API endpoint:

text
POST /flow_runs/{flow_run_id}/set_state

Example body:

json
{
  "state": {
    "type": "SCHEDULED",
    "name": "Scheduled",
    "message": "Retrying"
  },
  "force": true
}

Important caveats: - This is effectively retrying the entire flow run, not just failed tasks. - In practice, retries make sense for runs in terminal states like

FAILED

,

CRASHED

,

CANCELLED

, or

COMPLETED

. - If the run came from a deployment, a worker should pick it up after it is rescheduled. - If it did not come from a deployment, rescheduling alone may not be enough to actually execute it. There’s also a built-in CLI for this:

bash
prefect flow-run retry <flow-run-id>

I verified that command exists in Prefect 3.x. Useful references: - CLI help for flow-run retry: run

prefect flow-run retry --help

- Prefect orchestration client docs/source are reflected in

client.set_flow_run_state(...)

If you want, I can also give you: 1. a sync Python example, or 2. a raw

curl

example against the REST API.