A quick heads-up: in Prefect 3.x there is no

prefect flow delete

CLI command. So if “the CLI” is failing, it’s likely either: - a custom API call/script, or - a command aimed at something else like a deployment or flow run If deletion fails both in the UI and via API/client, that usually points to a backend issue rather than a UI-only problem. What should normally happen - Deleting a flow uses the API endpoint

DELETE /flows/{id}

- Prefect should cascade-delete related deployments and flow runs automatically - Existing deployments/runs generally should not block flow deletion Common causes when flow deletion fails in both places - The backend/API is erroring during the delete - The database transaction is timing out, especially if that flow has many related runs - A server/database integrity issue - Cloud/org permission issues if the token/user lacks sufficient access - A stale/bad flow ID, though that usually gives a straightforward not-found error Important note Deployments and flow runs are *not normally blockers*; they are expected to be deleted along with the flow. Best next step: try deleting via the Python client and capture the exact error

from prefect import get_client
import asyncio
from uuid import UUID

FLOW_ID = "your-flow-id-here"

async def main():
    async with get_client() as client:
        await client.delete_flow(UUID(FLOW_ID))
        print("Deleted")

asyncio.run(main())

If that fails, the exact traceback/error message will tell us a lot. Also useful checks 1. Confirm the flow still exists:

from prefect import get_client
import asyncio
from uuid import UUID

FLOW_ID = "your-flow-id-here"

async def main():
    async with get_client() as client:
        flow = await client.read_flow(UUID(FLOW_ID))
        print(flow)

asyncio.run(main())

2. If this is self-hosted Prefect Server, check the API/server logs at the time of deletion. 3. If the flow has a huge number of runs, try first deleting old flow runs or associated deployments, then retry the flow deletion. Useful references - Prefect Python client / flows API behavior is implemented in the Prefect server and client codebase: Prefect repository - Prefect Cloud UI: app.prefect.cloud Can you share: - whether you’re on Prefect Cloud or self-hosted Server - the exact command or method you used for the “CLI” attempt - the exact error message from the UI/API/CLI - whether this flow has a very large run history With that, I can help narrow it down quickly.

Thanks — this is helpful. A 500 from

DELETE /flows/{id}

means the problem is on the Prefect server side, not in your script. Because you’re using a self-hosted API at

<http://prefect.k8s.transmetrics.eu|prefect.k8s.transmetrics.eu>

, the next thing to check is the API server logs at the moment the delete request happens. Most likely causes in Prefect 3.x 1. An exception during event emission Prefect emits a

flow.deleted

event before finishing the delete. If event publishing fails, the whole request can return 500. 2. A slow or failing cascade delete If the flow has lots of related flow runs / states / deployments, the delete transaction can fail or time out. 3. Database/connection-pool issues Locking, pool exhaustion, or DB errors can also surface as 500s. What to do next 1. Check the Prefect API server logs Look for the traceback around the time you run the delete. If you’re on Kubernetes, something like:

kubectl logs <prefect-api-pod> -n <namespace> --since=10m

Or if multiple pods:

kubectl logs deployment/<prefect-api-deployment> -n <namespace> --since=10m

You want the full traceback for the

DELETE /api/flows/ad7112b3-a0ff-45ae-b7d0-32852538e46a

request. Search for terms like: -

emit

-

event

-

rollback

-

ForeignKey

-

constraint

-

TimeoutError

-

TooManyConnections

-

deadlock

2. Check how large this flow is If this flow has a huge run history, deletion may be choking on the cascade. You can test by listing the deployments and runs for that flow first. If you want, I can help you with a short script for that. 3. Try deleting deployments first Even though deployments should cascade automatically, manually removing them can reduce the amount of work in the final flow delete. If you know the deployments for that flow, try deleting them first, then retry the flow deletion. 4. If this is Postgres-backed, inspect DB locks / activity If you have DB access, check whether the delete is being blocked or timing out. For Postgres:

SELECT pid, state, wait_event_type, wait_event, query
FROM pg_stat_activity
WHERE state <> 'idle';

And count runs for that flow:

SELECT COUNT(*) FROM flow_run
WHERE flow_id = 'ad7112b3-a0ff-45ae-b7d0-32852538e46a';

Why I think server logs are the key here Your client traceback only shows: - request reached the API - API returned 500 - no client-side issue So the real cause is almost certainly in the server traceback. One more useful diagnostic: confirm whether all deletes fail or just this flow Try deleting: - a small test flow, or - a deployment / flow run associated with this flow That helps distinguish: - global server problem → all deletes fail - data-specific problem → only this flow fails Relevant references - Prefect repo: github.com/PrefectHQ/prefect - Prefect Cloud app: app.prefect.cloud If you can paste the API server traceback from the logs, I can help pinpoint the exact failure very quickly.