Short answer: don’t run the worker on your laptop. Run

prefect worker start

on always‑on compute (EC2, an ECS service, Kubernetes, or a Dockerized service on a server) so it keeps polling your work pool when your machine is off. Also, small correction: ECR is the image registry; Prefect’s pool type is ECS. Did you mean an ECS work pool? Good options to keep your worker alive - EC2 instance (systemd service) - Pros: simple, inexpensive, easy to manage - You’ll need Prefect + AWS credentials on the instance - Example systemd unit:

# /etc/systemd/system/prefect-worker.service
    [Unit]
    Description=Prefect Worker for ECS pool
    After=network-online.target

    [Service]
    Environment=PREFECT_API_URL=<https://api.prefect.cloud/api/accounts/<account-id>/workspaces/<workspace-id>>
    Environment=PREFECT_API_KEY=<your-api-key>
    # Provide AWS creds via instance role (preferred) or env vars if needed
    ExecStart=/usr/local/bin/prefect worker start -p my-ecs-pool -n ec2-worker-1
    Restart=always
    RestartSec=5

    [Install]
    WantedBy=multi-user.target

Then:

sudo systemctl daemon-reload
    sudo systemctl enable --now prefect-worker

- Docker on a server (with restart policy) - Pros: easy to ship/run; Docker restarts it on reboot - Example compose:

services:
      worker:
        image: prefecthq/prefect:3
        command: prefect worker start -p my-ecs-pool -n docker-worker-1
        environment:
          PREFECT_API_URL: <https://api.prefect.cloud/api/accounts/<account-id>/workspaces/<workspace-id>>
          PREFECT_API_KEY: <your-api-key>
          # If not using instance/task roles:
          # AWS_REGION: <region>
          # AWS_ACCESS_KEY_ID: <key>
          # AWS_SECRET_ACCESS_KEY: <secret>
        restart: unless-stopped

- ECS service (Fargate or EC2) - Build a small container whose entrypoint is:

prefect worker start -p my-ecs-pool -n ecs-service-worker

- Push it to ECR, then run it as an ECS Service with desired count 1 and a task role that allows ECS run-task, logs, etc. This keeps a worker alive in ECS itself. - Kubernetes - Run a Deployment with a Prefect image and the command above; Kubernetes will restart the pod if it dies. Key notes - In all cases, configure your Prefect connection via

PREFECT_API_URL

and

PREFECT_API_KEY

, and give the worker AWS permissions (ideally via an instance or task role) so it can start ECS tasks for your flow runs. - Command you’ll run in these environments:

prefect worker start -p my-ecs-pool

You can add

-n <name>

to identify the worker. Alternative: avoid a worker entirely - If you’d prefer not to run a long‑lived worker, you can use a serverless/push work pool type. Push pools do not require a worker; Prefect triggers your runs directly in the provider. If that’s appealing, I can help you pick the right push pool for AWS and set it up. Helpful docs - ECS work pool guide: Prefect AWS ECS guide - Prefect settings (API URL and key): Manage Prefect settings - Running long‑lived processes: Daemonize processes - Serverless/push pools overview: Serverless (push) work pools A couple of quick questions so I can tailor exact steps: - Can you confirm you’re using an ECS work pool (not ECR)? - Where would you prefer to host the worker: EC2, ECS service, Kubernetes, or a Dockerized server? - Are you using Prefect Cloud, and do you already have an API key handy?

You’ll want to run

prefect worker start

on always‑on compute, not your laptop. For an ECS work pool, good options are: - EC2 instance (as a systemd service) - Keeps the worker running across reboots and restarts. - Provide Prefect API settings and AWS permissions (ideally via an instance role). - Example:

# /etc/systemd/system/prefect-worker.service
    [Unit]
    Description=Prefect Worker for ECS pool
    After=network-online.target

    [Service]
    Environment=PREFECT_API_URL=<https://api.prefect.cloud/api/accounts/<account-id>/workspaces/<workspace-id>>
    Environment=PREFECT_API_KEY=<your-api-key>
    ExecStart=/usr/local/bin/prefect worker start -p my-ecs-pool -n ec2-worker-1
    Restart=always
    RestartSec=5

    [Install]
    WantedBy=multi-user.target

Then:

sudo systemctl daemon-reload
    sudo systemctl enable --now prefect-worker

- Docker on a server (with restart policy) - Easy to run anywhere you have Docker; Docker restarts it on reboot/crash. - Example:

services:
      worker:
        image: prefecthq/prefect:3
        command: prefect worker start -p my-ecs-pool -n docker-worker-1
        environment:
          PREFECT_API_URL: <https://api.prefect.cloud/api/accounts/<account-id>/workspaces/<workspace-id>>
          PREFECT_API_KEY: <your-api-key>
          # Prefer instance/task roles for AWS; env vars only if needed:
          # AWS_REGION: <region>
          # AWS_ACCESS_KEY_ID: <key>
          # AWS_SECRET_ACCESS_KEY: <secret>
        restart: unless-stopped

- ECS service (runs the worker inside ECS) - Build or use a Prefect image and set the command to:

prefect worker start -p my-ecs-pool -n ecs-service-worker

- Run it as an ECS Service (desired count 1) with a task role that can create ECS tasks, write logs, etc. This keeps the worker alive inside ECS itself. - Kubernetes - Run a Deployment that starts the worker; K8s will keep it healthy. Notes - Make sure the worker can reach Prefect Cloud (

PREFECT_API_URL

and

PREFECT_API_KEY

) and has AWS permissions (instance/task role recommended). - Command you’ll run in these environments:

prefect worker start -p my-ecs-pool

Optionally add

-n <name>

and

--work-queue <queue>

. If you’d prefer not to run a long‑lived worker at all, consider switching to a serverless/push work pool type for AWS; push pools don’t require a worker. Helpful docs - ECS work pool guide: Prefect AWS ECS guide - Prefect settings (API URL and key): Manage Prefect settings - Daemonizing workers (systemd, etc.): Daemonize processes - Serverless/push pools overview: Serverless (push) work pools Want me to tailor an exact setup? Tell me which target you prefer (EC2, ECS service, Docker on a server, or Kubernetes) and I’ll provide the step‑by‑step.