This note is about how I deployed airflow helm chart on home k3s cluster.
The content would be:

• Build custom airflow docker image to use KubernetesPodOperator
• Value settings for git sync
• Configure kubernetes connection

Build Airflow Image

I just made docker file like follows:

FROM apache/airflow:2.10.5-python3.11

USER airflow

RUN pip install apache-airflow-providers-cncf-kubernetes==10.5.0

and pushed it to my custom harbor repository

docker aa.bb.cc
docker build --platform linux/amd64 -t aa.bb.cc/common/airflow:latest -f Dockerfile .
docker push aa.bb.cc/common/airflow:latest

To use the image for the airflow, we need to change the values of the helm chart.
I changed two parts like follows:

defaultAirflowRepository: aa.bb.cc/common/airflow

registry:
  connection:
    user: ~
    pass: ~
    host: aa.bb.cc
    email: ~

Value Settings for Git Sync

To use git repository for gettting dag files, we need to set secret and corresponding values.
I decided to use ssh login for git sync, so I made secret like follows:

apiVersion: v1
kind: Secret
metadata:
  name: airflow-git-ssh-secret
  namespace: common
type: Opaque
data:
  # key needs to be gitSshKey
  gitSshKey: ~~~~~

To get the gitSshKey value, I used ssh-keygen and encoded with base64.
The important things is that it should include all text(like "--- begin …").

To set values, you can following settings:

executor: "KubernetesExecutor"

scheduler:
  extraVolumes:
    - name: git-sync-ssh-key
      secret:
        secretName: airflow-git-ssh-secret

dags:
  gitSync:
    enabled: true

    repo: git@github.com:rst0070/example.git
    branch: main # I use main branch
    rev: HEAD

    ref: main
    subPath: "dags" # I store dag files inside "dags" folder
    sshKeySecret: airflow-git-ssh-secret # I named the secret as airflow-git-secret

You can refer to this issue, If you are curious about why the scheduler.extraVolumes is required.

Configure kubernetes connection

I followed instruction from astronomer.
So the connection setting is following:

{
    "apiVersion": "v1",
    "clusters": [
        {
            "cluster": {
                "certificate-authority-data": "<certificate-authority-data>",
                "server": "https://kubernetes.docker.internal:6443"
            },
            "name": "docker-desktop"
        }
    ],
    "contexts": [
        {
            "context": {
                "cluster": "docker-desktop",
                "user": "docker-desktop"
            },
            "name": "docker-desktop"
        }
    ],
    "current-context": "docker-desktop",
    "kind": "Config",
    "preferences": {},
    "users": [
        {
            "name": "docker-desktop",
            "user": {
                "client-certificate-data": "<client-certificate-data>",
                "client-key-data": "<client-key-data>"
            }
        }
    ]
}

As you can see the configuration has same structure of kubectl config file.