< Marvin> my flow seems to run with values that are no longe Prefect Community #ask-marvin

<@ULVA73B9P> my flow seems to run with values tha...

Roger Sousa

09/24/2025, 8:56 PM

@Marvin my flow seems to run with values that are no longer in the code. I serve a deployment with multiple Cron schedules. (Tasks omitted for brevity.) from prefect import flow from prefect.server.schemas.schedules import Cron @flow def my_flow(list_schools, max_run_time, url_base): df = task_A(list_schools, max_run_time, url_base) task_B(df) if name == "__main__": URL_BASE = "www.sitesdasescolas.com.br" public_schools = ["2222-1111","2222-2222","2222-3333","2222-4444"] private_schools = ["3333-1111","3333-2222","3333-3333","3333-4444"] philanthropic_schools = ["5555-1111","5555-2222","5555-3333","5555-4444"] schedules = [ Cron( "0 9,12,15,18,21 * * *", slug="public", timezone="America/Sao_Paulo", parameters={ "list_schools": public_schools, "max_run_time": 180, "url_base": URL_BASE, }, ), Cron( "0 9,17 * * *", slug="private", timezone="America/Sao_Paulo", parameters={ # NOTE: this might be a bug — see below "deployment_ids": private_schools, "max_run_time": 480, "url_base": URL_BASE, }, ), Cron( "0 9,12,15,18,21,23 * * *", slug="philanthropic", timezone="America/Sao_Paulo", parameters={ # NOTE: this might be a bug — see below "deployment_ids": philanthropic_schools, "max_run_time": 180, "url_base": URL_BASE, }, ), ] my_flow.serve( name="Flow escolas", schedules=schedules, version="1.0.0", pause_on_shutdown=False, tags=["escolas"], ) During the flow’s execution, I noticed the school codes in the private_schools list were wrong. I changed the list values to: private_schools = ["9999-1111","9999-2222","9999-3333","9999-4444"] Then I served/redeployed the flow again. When the scheduled run triggers, it still uses the old private values (the 3333-... list), not the updated 9999-... list. What might explain this behavior and how can I fix it?

Marvin

09/24/2025, 8:56 PM

✅ thought for 442.1 seconds

Marvin

09/24/2025, 9:03 PM

Short answer: you’re seeing “stale” parameters because Prefect pre-creates scheduled flow runs. Those runs keep the parameters they had at the time they were created. Re-serving the deployment updates the schedule going forward, but any already-created runs will still use the old values. Two additional notes from your snippet: - In your

private

and

philanthropic

schedules you passed

deployment_ids

instead of

list_schools

. Your flow signature is

my_flow(list_schools, max_run_time, url_base)

, so those runs would not receive

list_schools

unless you fix that key. - In Prefect 3.x, schedule parameters are stored on the schedule and copied onto each created flow run. They’re not dynamically read from your code at trigger time. What to do now 1) Fix the parameter key and re-serve

Copy code

schedules = [
       Cron(
           "0 9,17 * * *",
           slug="private",
           timezone="America/Sao_Paulo",
           parameters={
               "list_schools": private_schools,  # <-- fix
               "max_run_time": 480,
               "url_base": URL_BASE,
           },
       ),
       Cron(
           "0 9,12,15,18,21,23 * * *",
           slug="philanthropic",
           timezone="America/Sao_Paulo",
           parameters={
               "list_schools": philanthropic_schools,  # <-- fix
               "max_run_time": 180,
               "url_base": URL_BASE,
           },
       ),
   ]
   my_flow.serve(name="Flow escolas", schedules=schedules, version="1.0.1", tags=["escolas"], pause_on_shutdown=False)

2) Clear or cancel any pre-created runs that still hold the old parameters - In the UI: go to your deployment → Scheduled tab → cancel/delete the old future runs. - Via CLI: - List scheduled runs:

Copy code

prefect flow-run ls --flow-name "my_flow" --state SCHEDULED --limit 200

- Cancel or delete each affected run:

Copy code

prefect flow-run cancel <RUN_ID>

Copy code

prefect flow-run delete <RUN_ID>

- You can inspect a run to confirm its parameters:

Copy code

prefect flow-run inspect <RUN_ID>

3) Optionally, reset the schedules to avoid confusion - List schedules on the deployment and grab the schedule IDs:

Copy code

prefect deployment schedule ls "Flow escolas"

- Pause or delete the old schedule (e.g., the “private” one), then re-serve (or create a new schedule with a new slug like

private-v2

Copy code

prefect deployment schedule pause "Flow escolas" <SCHEDULE_ID>

Copy code

prefect deployment schedule delete "Flow escolas" <SCHEDULE_ID>

Tips to avoid this next time - After changing schedule parameters, either change the schedule slug (to force a fresh schedule) or cancel/delete any pre-created “Scheduled” runs so only new runs (with updated parameters) remain. - Keep

enforce_parameter_schema=True

(the default for

serve

) so typos like

deployment_ids

list_schools

are caught early. Docs for reference: - Schedules (Prefect 3) - Python schedule API

Open in Slack

Previous Next