@Marvin, I want to override specific job configurations (cpu and memory pod defs) for specific prefect deployments (running on Kubernetes) for example, my work pool is defined like:

...
  "job_configuration": {
...
    "job_manifest": {
      "kind": "Job",
      "spec": {
        "template": {
          "spec": {
            "containers": [
              {
                "env": "{{ env }}",
                "args": "{{ command }}",
                "name": "prefect-job",
                "image": "{{ image }}",
                "resources": {
                  "limits": {
                    "cpu": "1000m",
                    "memory": "6G",
                    "ephemeral-storage": "1Gi"
                  },
                  "requests": {
                    "cpu": "1000m",
                    "memory": "6G",
                    "ephemeral-storage": "1Gi"
                  }
...

I know that we can set "Job Variables" for deployment, but im not sure if thats the approach and how to set it for example to have "memory": 60G I have this in my deployment config - but it's affects nothing

{
  "resources": {
    "limits": {
      "cpu": "3000m",
      "memory": "50G",
      "ephemeral-storage": "1Gi"
    },
    "requests": {
      "cpu": "3000m",
      "memory": "50G",
      "ephemeral-storage": "1Gi"
    }
  }
}

To override specific job configurations like CPU and memory for Prefect deployments running on Kubernetes, you can use the

job_variables

argument in your deployment configuration. This allows you to pass a dictionary of variables that can override any values set on the work pool or deployment. Here's a concise guide on how to do this: 1. Define Job Variables: You need to add the desired resource configurations, such as CPU and memory limits, in the

job_variables

argument when deploying your flow. 2. Deployment Configuration: Ensure that your deployment script includes the

job_variables

dictionary with the appropriate resource settings. This can be done in your deployment script or via the Prefect UI. 3. Example Usage: In your deployment configuration, you might include something like:

python
   deployment = Deployment(
       flow=my_flow,
       name="my_deployment",
       work_pool_name="my_k8s_work_pool",
       job_variables={
           "resources": {
               "limits": {
                   "cpu": "3000m",
                   "memory": "60G",
                   "ephemeral-storage": "1Gi"
               },
               "requests": {
                   "cpu": "3000m",
                   "memory": "60G",
                   "ephemeral-storage": "1Gi"
               }
           }
       }
   )

4. UI and CLI: You can also use the Prefect UI or CLI to pass these job variables during a custom run, allowing you to override the default settings for specific runs. For more detailed guidance, you may refer to the following resources: - Kubernetes Infrastructure Examples - Customize Infrastructure Remember that Prefect 3.x has introduced changes such as replacing agents with workers and infrastructure blocks with work pools, so ensure your configurations align with these updates.