Bring your towel and join one of the fastest growing data communities. Welcome to our second-generation open source orchestration platform, a completely rethought approach to dataflow automation.

Prefect Community

I'm currently attempting to configure our flow run deployments such that they can override the `ECSTask` infra block by specifying custom values for `cpu` and `memory`.

Any help is appreciated!

We use `Deployment.build_from_flow` to create our deployments programmatically:
• `flow` is the function that defines the flow logic
• `infrastructure` is set to the `ECSTask` block
I'd like to know the most reliable manner (a simple code example) to pass parameters to `Deployment.build_from_flow` (is it `infra_overrides`?) or modify the `ECSTask` block (without changing the default settings on the block).

Ideally, I'm hoping for something like `infra_overrides={"cpu": 1024, "memory": 4096}` .

Also, I'm curious about whether this would register a new task definition every time the deployment is run.

Hi Kelvin -
Infra overrides expects a dict , I know specifically for kubernetes jobs you can pass an list called customizations with the resource parameters - I’d need to double check syntax for ECSTask

Hey <@U03FKQX1TH9> , thanks for the reply! Do let me know about the right syntax!

<https://prefecthq.github.io/prefect-aws/ecs/>
```Run a task with custom memory and CPU requirements


ECSTask(command=["echo", "hello world"], memory=4096, cpu=2048)```

Alternatively:
```Run a task with custom VPC subnets


ECSTask(
    command=["echo", "hello world"],
    task_customizations=[
        {
            "op": "add",
            "path": "/networkConfiguration/awsvpcConfiguration/subnets",
            "value": ["subnet-80b6fbcd", "subnet-42a6fdgd"],
        },
    ]
)```

task customizations accept a json patch, which you can treat funcitonally the same as you would for a kubernetes resource block

Hey <@U03FKQX1TH9>

Thanks for the reply!

Our situation is a bit different. We have an ECS Task definition already and we reference it by its ARN.
Could we still apply these `task_customizations` ? Also, how would the `task_customizations` look, if we wanted to specify a certain amount for `cpu` / `memory`?