Deploying agents to the Flyte sandbox¶

After you have finished testing an agent locally, you can deploy your agent to the Flyte sandbox.

Here’s a step by step guide to deploying your agent image to the Flyte sandbox.

Start the Flyte sandbox:

flytectl demo start

Build an agent image: You can go to here to see the Dockerfile we use in flytekit python. Take Databricks agent as an example:

FROM python:3.9-slim-bookworm

RUN apt-get update && apt-get install build-essential git -y
RUN pip install prometheus-client grpcio-health-checking
RUN pip install --no-cache-dir -U flytekit \
    git+https://github.com/flyteorg/flytekit.git@<gitsha>#subdirectory=plugins/flytekit-spark \
    && apt-get clean autoclean \
    && apt-get autoremove --yes \
    && rm -rf /var/lib/{apt,dpkg,cache,log}/ \
    && :

CMD pyflyte serve agent --port 8000

docker buildx build -t localhost:30000/flyteagent:example -f Dockerfile.agent . --load
docker push localhost:30000/flyteagent:example

Deploy your agent image to the Kubernetes cluster:

kubectl edit deployment flyteagent -n flyte

Search for the image key and change its value to your agent image:

image: localhost:30000/flyteagent:example

Set up your secrets: Let’s take Databricks agent as an example:

kubectl edit secret flyteagent -n flyte

Get your BASE64_ENCODED_DATABRICKS_TOKEN:

echo -n "<DATABRICKS_TOKEN>" | base64

Add your token to the data field:

apiVersion: v1
data:
  flyte_databricks_access_token: <BASE64_ENCODED_DATABRICKS_TOKEN>
kind: Secret
metadata:
  annotations:
    meta.helm.sh/release-name: flyteagent
    meta.helm.sh/release-namespace: flyte
  creationTimestamp: "2023-10-04T04:09:03Z"
  labels:
    app.kubernetes.io/managed-by: Helm
  name: flyteagent
  namespace: flyte
  resourceVersion: "753"
  uid: 5ac1e1b6-2a4c-4e26-9001-d4ba72c39e54
type: Opaque

Note

Please ensure two things:

The secret name consists only of lowercase English letters.
The secret value is encoded in Base64.

Restart development:

kubectl rollout restart deployment flyte-sandbox -n flyte

Test your agent remotely in the Flyte sandbox:

pyflyte run --remote agent_workflow.py agent_task

Note

You must build an image that includes the plugin for the task and specify its config with the --image flag when running pyflyte run or in an ImageSpec definition in your workflow file.