Install Official Airflow chart using helm 3 "Google Kubernetes Engine"

Add repo airflow

helm repo add apache-airflow https://airflow.apache.org
helm repo update

helm repo list
NAME          	URL                       
apache-airflow	https://airflow.apache.org

helm search repo airflow --versions
NAME                  	CHART VERSION	APP VERSION	DESCRIPTION                                       
apache-airflow/airflow	1.0.0        	2.0.2      	Helm chart to deploy Apache Airflow, a platform...

If you need get new value

helm show values apache-airflow/airflow --version=1.0.0 > values.yaml

Create namespace

kubectl create namespace airflow

Apply configmap variables

kubectl apply -f variables.yaml

Install airflow

Airflow 2.0 allows users to run multiple schedulers. This feature is only recommended for PostgreSQL

• values.yaml will replaced by --set config
• executor = Airflow executor
• flower.enabled = Enable Flower (web based tool for monitoring and administrating Celery)
• webserver.service.type= you can change from ClusterIP to LoadBalancer
• --version=1.0.0 = prevent updated helm repo versions and make errors

helm install airflow apache-airflow/airflow --namespace airflow -f values.yaml --version=1.0.0 \
--set airflowVersion=2.0.2 \
--set defaultAirflowTag=2.0.2 \
--set executor=CeleryExecutor \
--set redis.enabled=True \
--set flower.enabled=True \
--set webserver.service.type=ClusterIP \
--set webserver.replicas=2 \
--set scheduler.replicas=2 \
--set pgbouncer.enabled=True \
--set pgbouncer.maxClientConn=150 \
--set pgbouncer.metadataPoolSize=10 \
--set pgbouncer.resultBackendPoolSize=5 \
--set workers.replicas=5 \
--set workers.persistence.enabled=True \
--set workers.terminationGracePeriodSeconds=600 \
--set workers.persistence.size=10Gi

Login to UI airflow and flower (if you use ClusterIP via port-forward) user: admin pass:admin

kubectl port-forward svc/airflow-webserver 8080:8080 --namespace airflow
kubectl port-forward svc/airflow-flower 5555:5555 --namespace airflow

Dockerfile and DAGS structure

.
├── dags
│   ├── example
│   │   └── example.py
│   └── generated
│       └── example-generated.py
└── Dockerfile
    requirements.txt
    values.yaml
    variables.yaml

Deploy new DAG

Build docker image

docker build --no-cache -t zeus-airflow .
docker tag zeus-airflow asia.gcr.io/zeus-cloud/zeus-airflow:0.6
docker push asia.gcr.io/zeus-cloud/zeus-airflow:0.6

Apply configmap variables

kubectl apply -f variables.yaml

Update airflow with new docker image and dag + auto rollback

• --atomic = if set, upgrade process rolls back changes made in case of failed upgrade
• --timeout = deployment timeout if more than 360s
• workers.replicas= number pods worker for scale out running jobs
• images.airflow.tag=0.6 (docker tag)

helm upgrade --install --atomic --timeout 360s airflow apache-airflow/airflow --namespace airflow -f values.yaml --version=1.0.0 \
--set airflowVersion=2.0.2 \
--set defaultAirflowTag=2.0.2 \
--set executor=CeleryExecutor \
--set redis.enabled=True \
--set flower.enabled=True \
--set webserver.service.type=ClusterIP \
--set webserver.replicas=2 \
--set scheduler.replicas=2 \
--set pgbouncer.enabled=True \
--set pgbouncer.maxClientConn=150 \
--set pgbouncer.metadataPoolSize=10 \
--set pgbouncer.resultBackendPoolSize=5 \
--set workers.replicas=4 \
--set workers.persistence.enabled=True \
--set workers.persistence.size=10Gi \
--set images.airflow.repository=asia.gcr.io/zeus-cloud/zeus-airflow \
--set images.airflow.tag=0.6 \
--set images.airflow.pullPolicy=IfNotPresent

Manual Rollback

helm history airflow --namespace airflow
REVISION	UPDATED                 	STATUS          	CHART        	APP VERSION	DESCRIPTION                                                  
1       	Fri Jun 11 11:58:19 2021	superseded      	airflow-1.0.0	2.0.2      	Install complete                                             
2       	Fri Jun 11 12:05:28 2021	superseded      	airflow-1.0.0	2.0.2      	Upgrade complete                                             
3       	Fri Jun 11 12:09:09 2021	pending-upgrade 	airflow-1.0.0	2.0.2      	Preparing upgrade                                            
4       	Fri Jun 11 12:15:30 2021	deployed        	airflow-1.0.0	2.0.2      	Rollback to 2                                                
5       	Fri Jun 11 12:22:32 2021	failed          	airflow-1.0.0	2.0.2      	Upgrade "airflow" failed: timed out


helm rollback airflow 4 --namespace airflow
Rollback was a success! Happy Helming!

You can get Fernet Key value by running the following:

echo Fernet Key: $(kubectl get secret --namespace airflow airflow-fernet-key -o jsonpath="{.data.fernet-key}" | base64 --decode)

Sometime if deployment error and helm auto rollback and pods worker still ImagePullBackOff

kubectl delete pods -n airflow airflow-worker-0
pod "airflow-worker-0" deleted

Change Worker cpu and memory or nodepool

Edit values.yaml and change all nodeSelector

  resources:
    limits:
     cpu: 1600m
     memory: 1800Mi
    requests:
     cpu: 1200m
     memory: 1400Mi

  nodeSelector:
    cloud.google.com/gke-nodepool: pool-airflow
  affinity: {}
  tolerations: []

Example useful airflow helm and cli

airflow config list
helm ls -n airflow

Uninstall airflow

helm delete airflow --namespace airflow

Reference

https://eatcodeplay.com/a-simple-dag-to-quickly-purge-old-airflow-logs-274ed5de1567 (delete logs with dag)
https://airflow.apache.org/docs/helm-chart/stable/parameters-ref.html
https://airflow.apache.org/docs/helm-chart/stable/index.html