A few likely culprits when deployments don’t show up in the UI. Quick checklist (assuming Prefect 3.x): 1) Your container may not actually be pointing at the right API - Inside the running container, check what Prefect thinks the API is:

prefect config view --show-sources

- Make sure

PREFECT_API_URL

is reachable from inside the container and includes the

/api

suffix. - Don’t use

<http://localhost>:...

unless the Prefect server is in the same container. In Docker, “localhost” is the container itself. Use the service hostname or an external URL, e.g.

<http://prefect-server:4200/api>

or

<https://prefect.example.com/api>

. 2) Watch the logs when

serve

starts -

prefect.serve

will create/update deployments at startup. You should see log lines like “Created deployment …” or clear connection errors. - If the API is unreachable or misconfigured, Prefect will complain loudly. If it’s suspiciously quiet, the env var may not be set in that container. 3) Verify versions match - SDK and Server should both be 3.x. Mismatched major versions won’t talk nicely.

prefect version

Check the server’s version in the UI footer, and the SDK’s in the container. 4) Ensure a work pool exists (and is the one you’re targeting) - By default, deployments go to the “default” process work pool. If that pool doesn’t exist on your server, creation will fail. - You can set it explicitly to be clear:

flow_1 = my_flow.to_deployment(
    name="Healthcheck Flow",
    description="A simple flow that says hello.",
    parameters={"name": "Nathan"},
    cron="*/5 * * * *",
    tags=["test"],
    work_pool_name="default",
)
flow_2 = my_flow_2.to_deployment(
    name="Healthcheck Flow 2",
    description="A simple flow that says hello.",
    cron="* * * * *",
    tags=["test"],
    work_pool_name="default",
)
serve(flow_1, flow_2)

- Even if there’s no worker attached to the pool yet, the deployments should still appear in the UI once registration succeeds. 5) Sanity-check connectivity from inside the container - Quick probe to the API health endpoint:

python - <<'PY'
import os, requests
url = os.environ.get("PREFECT_API_URL", "")
print("PREFECT_API_URL =", url)
url = url.rstrip("/")
resp = requests.get(f"{url}/health", timeout=5)
print("GET /health ->", resp.status_code, resp.text[:200])
PY

If this doesn’t surface the issue, could you share: - The exact value you’re using for

PREFECT_API_URL

(redact host if needed; format matters, especially the

/api

). - The output of

prefect config view --show-sources

from inside the container. - Any logs printed when the container starts (particularly around “serve” and deployment creation). - SDK and server versions. With those, we can narrow it down quickly. And yes, I know, computers should just read our minds by now. Sadly they still insist on URLs that aren’t localhost.

Thanks — the API URL looks correct from inside the container. Since your deployments register fine but nothing is running on the schedule, there are two prime suspects in Prefect 3.x: 1) The deployment is paused - New deployments can be created in a paused state. Let’s confirm:

uv run prefect deployments inspect "my_flow/Healthcheck Flow"

Look for “paused” and the “schedules” block. - If paused, unpause it in the UI or make it explicit in code:

flow_1 = my_flow.to_deployment(
    name="Healthcheck Flow",
    parameters={"name": "Nathan"},
    cron="*/5 * * * *",
    tags=["test"],
    paused=False,
)

(Same for

flow_2

.) 2) The server scheduler is not running - If your self-hosted server was started with services disabled (or the scheduler service off), the UI will show deployments but no scheduled runs will be created. - Ensure your server starts with the scheduler enabled, e.g. don’t use

--no-services

, and do use: - CLI:

prefect server start --scheduler

(or just no flag; scheduler should be on by default) - Env: set

PREFECT_API_SERVICES_SCHEDULER=true

- There’s also a

--late-runs/--no-late-runs

toggle; it only affects catching up missed runs, not creating new on-time runs. Quick triage steps - See if the server created any runs at all:

uv run prefect deployments ls
uv run prefect deployments inspect "my_flow/Healthcheck Flow"

- Try a manual run to confirm the runner can execute:

uv run prefect deployments run "my_flow/Healthcheck Flow"

If manual runs start but scheduled ones don’t appear, it’s almost certainly the scheduler service. If you can share how you start Prefect Server in Coolify (the exact command/flags or env vars), I can point to the precise fix. Otherwise, set

paused=False

in your

to_deployment(...)

calls and make sure your server is started with the scheduler enabled. If only everything obeyed cron as faithfully as we do.