Thanks for the details — a couple things to unpack here. First, to confirm assumptions: I’m assuming you’re on Prefect 3.x and that you’re running the webserver with

prefect server start --no-services

and a separate background-services pod with

prefect server services start

. If that’s not right, let me know. What usually causes “duplicate cron schedules” after

prefect deploy --all

- If you see multiple schedule entries under a single deployment (e.g., two identical cron rows), it’s typically because the schedules in

prefect.yaml

don’t have stable identifiers. On each deploy, Prefect can’t match the “same” schedule to update it, so it creates a new one. - If you see duplicate flow runs at the same timestamps (not duplicate schedule rows), it’s usually because more than one Scheduler service is running. How to make deploys idempotent for schedules - Give each schedule a stable slug in your

prefect.yaml

. Slugs let Prefect match and update an existing schedule instead of creating a new one on every deploy. Example

prefect.yaml

snippet:

deployments:
  - name: my-deployment
    entrypoint: flows/flow.py:my_flow
    schedules:
      - cron: "0 * * * *"
        timezone: "UTC"
        active: true
        slug: hourly-main

- One-time cleanup: clear existing duplicate schedules, then redeploy once with slugs in place:

prefect deployment schedule ls "<flow-name>/<deployment-name>"
prefect deployment schedule clear "<flow-name>/<deployment-name>" -y
prefect deploy --all

You can list/clear schedules with the CLI: -

prefect deployment schedule ls

and

prefect deployment schedule clear

are available; see

--help

for usage. - Verified commands: -

prefect deployment schedule --help

-

prefect deployment schedule clear --help

If you are also calling

prefect deployment schedule create

elsewhere (e.g., CI/CD), that will add additional schedules too. Either remove that step and rely solely on

prefect deploy --all

+ slugs, or make sure you clear/replace before creating. You can inspect options with:

prefect deployment schedule create --help

About the Scheduler in self-hosted setups - Ensure only one Scheduler service is running. If you accidentally have more than one instance (e.g., scaled your services pod >1 replicas, or also started services in another place), you can get duplicate flow runs at the same timestamp. - Your pattern of

prefect server start --no-services

+ a single

prefect server services start

process is correct; just verify you don’t have multiple replicas of the services pod. Helpful references - Schedules concept and YAML config: Schedules in Prefect 3 - Creating schedules for deployments: Create schedules -

prefect.yaml

deployment config: prefect.yaml reference - Deployment upsert behavior (note that scheduled runs are cleared on upsert): Create deployment API - Scheduler service: Scheduler service A couple quick questions to target the fix: 1) Are you seeing duplicate schedule entries under the deployment, or duplicate flow runs at the same times? 2) Could you share the

schedules:

block from your

prefect.yaml

(with any secrets removed)? 3) How many replicas of your background services pod are running? Just 1? If it’s duplicate schedule entries, adding stable

slug

values and doing a one-time

schedule clear

+ redeploy should resolve it. If it’s duplicate runs, we’ll focus on ensuring only one Scheduler instance is active.