Ask AI

Kubernetes agent configuration reference#

This guide is applicable to Dagster Cloud.

This reference describes the various configuration options Dagster Cloud currently supports for Kubernetes agents.


Viewing the Helm chart#

To see the different customizations that can be applied to the Kubernetes agent, you can view the chart's default values:

helm repo add dagster-cloud https://dagster-io.github.io/helm-user-cloud
helm repo update
helm show values dagster-cloud/dagster-cloud-agent

You can also view the chart values on ArtifactHub.


Per-deployment configuration#

The workspace value of the Helm chart provides the ability to add configuration for all jobs that are spun up by the agent, across all repositories. To add secrets or mounted volumes to all Kubernetes Pods, you can specify your desired configuration under this value.

Additionally, the imagePullSecrets value allows you to specify a list of secrets that should be included when pulling the images for your containers.


Per-location configuration#

When adding a code location to Dagster Cloud with a Kubernetes agent, you can use the container_context key on the location configuration to add additional Kubernetes-specific configuration. If you're using the Dagster Cloud Github action, the container_context key can also be set for each location in your dagster_cloud.yaml file, using the same format.

The following example dagster_cloud.yaml file illustrates the available fields:

# dagster_cloud.yaml

locations:
  - location_name: cloud-examples
    image: dagster/dagster-cloud-examples:latest
    code_source:
      package_name: dagster_cloud_examples
    container_context:
      k8s:
        env_config_maps:
          - my_config_map
        env_secrets:
          - my_secret
        env_vars:
          - FOO_ENV_VAR=foo_value
          - BAR_ENV_VAR
        image_pull_policy: Always
        image_pull_secrets:
          - name: my_image_pull_secret
        labels:
          my_label_key: my_label_value
        namespace: my_k8s_namespace
        service_account_name: my_service_account_name
        volume_mounts:
          - mount_path: /opt/dagster/test_mount_path/volume_mounted_file.yaml
            name: test-volume
            sub_path: volume_mounted_file.yaml
        volumes:
          - name: test-volume
            config_map:
              name: test-volume-configmap
        server_k8s_config: # Raw kubernetes config for code servers launched by the agent
          pod_spec_config:
            node_selector:
              disktype: standard
          container_config:
            resources:
              limits:
                cpu: 100m
                memory: 128Mi
        run_k8s_config: # Raw kubernetes config for runs launched by the agent
          pod_spec_config:
            node_selector:
              disktype: ssd
          container_config:
            resources:
              limits:
                cpu: 500m
                memory: 1024Mi

Environment variables and secrets#

Using the container_context.k8s.env_vars and container_context.k8s.env_secrets properties, you can specify environment variables and secrets for a specific code location. For example:

# dagster_cloud.yaml

location:
  - location_name: cloud-examples
    image: dagster/dagster-cloud-examples:latest
    code_source:
      package_name: dagster_cloud_examples
    container_context:
      k8s:
        env_vars:
          - database_name
          - database_username=hooli_testing
        env_secrets:
          - database_password
PropertyDescription
env_varsA list of environment variable names to inject into the job, formatted as KEY or KEY=VALUE. If only KEY is specified, the value will be pulled from the current process.
env_secretsA list of secret names, from which environment variables for a job are drawn using envFrom. Refer to the Kubernetes documentation for more info.

Refer to the following guides for more info about environment variables:


Op isolation#

By default, each Dagster job will run in its own Kubernetes pod, with each op running in its own subprocess within the pod.

You can also configure your Dagster job with the k8s_job_executor to run each op in its own Kubernetes pod. For example:

from dagster import job
from dagster_k8s import k8s_job_executor

@job(executor_def=k8s_job_executor)
def k8s_job():
    ...

Per-job and per-op configuration#

To add configuration to specific Dagster jobs, ops, or assets, use the dagster-k8s/config tag. For example, to specify that a job should have certain resource limits when it runs. Refer to Customizing your Kubernetes deployment for Dagster Open Source for more info.


Running as a non-root user#

Starting in 0.14.0, the provided docker.io/dagster/dagster-cloud-agent image offers a non-root user with id 1001. To run the agent with this user, you can specify the dagsterCloudAgent value in the Helm chart to be:

dagsterCloudAgent:
  podSecurityContext:
    runAsUser: 1001

We plan to make this user the default in a future release.


Grant AWS permissions#

You can provide your Dagster pods with permissions to assume an AWS IAM role using a Service Account. For example, you might do this to configure an S3 IO Manager.

  1. Create an IAM OIDC provider for your EKS cluster

  2. Create an IAM role and and attach IAM policies

  3. Update the Helm chart to associate the IAM role with a service account:

     serviceAccount:
     create: true
     annotations:
       eks.amazonaws.com/role-arn: "arn:aws:iam::1234567890:role/my_service_account_role"
    

This will allow your agent and the pods it creates to assume the my_service_account_role IAM role.