Awesome

Install

Have Helm 3 installed.

helm repo add windmill https://windmill-labs.github.io/windmill-helm-charts/
helm install mywindmill windmill/windmill -n windmill --create-namespace --values values.yaml

To update versions:

helm repo update windmill
helm upgrade mywindmill windmill/windmill -n windmill --values values.yaml

You do not need to provide a values.yaml to be able to test it on minikube. Follow the steps below.

When using a non super-user role for postgresql in databaseUrl

You will need to setup some required roles which would otherwise be done automatically when using a super-user role for the databaseUrl.

Follow those instructions

Test it on minikube

To make it work on a local minkube to test. Get the ip address of the ingress:

▶ kubectl get ingress -n windmill
NAME       CLASS    HOSTS                        ADDRESS        PORTS   AGE
windmill   <none>   windmill,windmill,windmill   192.168.49.2   80      13m

If not ip address is displayed, enable the ingress addon:

minikube addons enable ingress

Then modify /etc/hosts to match the baseDomain, by default 'windmill'.

E.g:

192.168.49.2   windmill

Then open your browser at http://windmill

Even if you setup oauth, login as** admin@windmill.dev **/ changeme to setup the instance & accounts and give yourself admin privileges.

Core Values

# windmill root values block
windmill:
  # domain as shown in browser, this is used together with `baseProtocol` as part of the BASE_URL environment variable in app and worker container and in the ingress resource, if enabled
  baseDomain: windmill
  baseProtocol: http
  # postgres URI, pods will crashloop if database is unreachable, sets DATABASE_URL environment variable in app and worker container
  databaseUrl: postgres://postgres:windmill@windmill-postgresql/windmill?sslmode=disable
  # replica for the application app
  appReplicas: 2
  # replicas for the workers, jobs are executed on the workers
  lspReplicas: 2
  workerGroups:
    # The default worker group is the one that will execute jobs with any taggs  except the native ones. Windmill has a default worker group configuration for it
    - name: "default"
      replicas: 3
      # -- Annotations to apply to the pods
      annotations: {}

      # -- Labels to apply to the pods
      labels: {}

      # -- Node selector to use for scheduling the pods
      nodeSelector: {}

      # -- Tolerations to apply to the pods
      tolerations: []

      # -- Affinity rules to apply to the pods
      affinity: {}

      # -- Resource limits and requests for the pods
      resources:
        requests:
          memory: "1028Mi"
          cpu: "500m"
        limits:
          memory: "2048Mi"
          cpu: "1000m"

      # -- Extra environment variables to apply to the pods
      extraEnv: []

      # -- Extra sidecar containers
      extraContainers: []

      # -- Mode for workers, defaults to "worker" - alternative "agent" requires Enterprise license
      mode: "worker"

    # Thenative worker group will only execute native jobs. Windmill has a default worker group configuration for it
    - name: "native"
      replicas: 4
      # -- Resource limits and requests for the pods
      resources:
        requests:
          memory: "128Mi"
          cpu: "100m"
        limits:
          memory: "256Mi"
          cpu: "200m"

      # -- Extra environment variables to apply to the pods
      extraEnv: []

      # -- Extra sidecar containers
      extraContainers: []

      # -- Mode for workers, defaults to "worker" - alternative "agent" requires Enterprise license
      mode: "worker"

    - name: "gpu"
      replicas: 0

  # Use those to override the tag or image used for the app and worker containers. Windmill uses the same image for both.
  # By default, if enterprise is enable, the image is set to ghcr.io/windmill-labs/windmill-ee, otherwise the image is set to ghcr.io/windmill-labs/windmill
  #tag: "mytag"
  #image: "ghcr.io/windmill-labs/windmill"

# enable postgres (bitnami) on kubernetes
postgresql:
  enabled: true

# enable minio (bitnami) on kubernetes
minio:
  enabled: false

# Configure Ingress
# ingress:
#   className: ""

# enable enterprise features
enterprise:
  # -- enable windmill enterprise, requires license key.
  enabled: false

Full Values

Local S3

The chart includes a Minio S3 distribution to demonstrate the usage of S3 as a resource in a vendor-agnostic environment like Kubernetes. The local Minio S3 service will be available to the Windmill workers through its Kubernetes service, which is set to "windmill-minio" by default. In the Resources page, you should create an S3 API Connection Object, and import it as a connection object to reduce code duplication between scripts. For the sake of this example, this stage is skipped. Below is an example of how to authenticate and use the provided local S3 distribution in a Python script running in Windmill:

from minio import Minio

def main():
    # Create a client with the MinIO server, its access key
    # and secret key.
    client = Minio(
        "windmill-minio", # Local Kubernetes Service
        access_key="windmill",
        secret_key="windmill",
    )

    # Make 'demo' bucket if not exist.
    found = client.bucket_exists("demo")
    if not found:
        client.make_bucket("demo")
    else:
        print("Bucket 'demo' already exists")

    with open('readme.txt', 'w') as f:
        f.write('Create a new text file!')

    client.fput_object(
        "demo", "readme.txt", "readme.txt",
    )

    print(
        "'readme.txt' is successfully uploaded as "
        "object 'readme.txt' to bucket 'demo'."
    )

Enterprise features

To use the enterprise version with the <license key> provided upon subscription, add the following to the values.yaml file:

enterprise:
  enabled: true

Then go to the superadmin settings -> instance settings -> license key and set your license key

S3 Cache

Enterprise users can use S3 storage for dependency caching for performance reasons at high scale (use only with #workers > 20). Cache is two way synced at regular intervals (10 minutes). To use it, the worker deployment requires access to an S3 bucket. There are several ways to do this:

1. On AWS (and EKS) , you can use a service account with IAM roles attached. See AWS docs - once you have a policy , you can create an account via eksctl for instance

   eksctl create iamserviceaccount --name serviceaccountname --namespace production --cluster windmill-cluster --role-name "iamrolename" \ --attach-policy-arn arn:aws:iam::12312315:policy/bucketpolicy --approve
   

2. Mount/attach a credentials file in /root/.aws/credentials of the worker deployment

3. Add environment variables for the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY, via kube secrets.

The sync relies on rclone and uses its methods of authentication to s3 per Rclone documentation

Then the values settings become:

enterprise:
  enabled: true
  enabledS3DistributedCache: true
  s3CacheBucket: mybucketname

Caveats

• Postgres is included for demo purposes, it is a stateful set with a small volume claim applied. If you want to host postgres in k8s, there are better ways, or offload it outside your k8s cluster. Postgres can be disabled entirely in the values.yaml file.
• The postgres user/pass is currently not a secret/encrypted

Kubernetes Hosting Tips

Ingress configuration

The helm chart does have an ingress configuration included. It's enabled by default. The ingress uses the windmill.baseDomain variable for its hostname configuration. Here are example configurations for a few cloud providers.

It configures the HTTP ingress for the app, lsp and multiplayer containers. The configuration (except for plain nginx ingress) also exposes the windmill app SMTP service for email triggers on a separate IP address/domain name. This is the IP address/domain name you need to point your MX/A records to, learn more here.

AWS ALB

windmill:
  baseDomain: "windmill.example.com"
  app:
    smtpService:
      enabled: true
      annotations:
        service.beta.kubernetes.io/aws-load-balancer-type: "external"
        service.beta.kubernetes.io/aws-load-balancer-nlb-target-type: "ip"
        service.beta.kubernetes.io/aws-load-balancer-scheme: "internet-facing"
        # # for static ip (more info on https://kubernetes-sigs.github.io/aws-load-balancer-controller/v2.7/guide/service/annotations/#eip-allocations):
        # service.beta.kubernetes.io/aws-load-balancer-eip-allocations: eipalloc-xxxxxxxxxxxxxxxxx,eipalloc-yyyyyyyyyyyyyyyyy
    ...
  ...
...

ingress:
  className: "alb"
  annotations:
    alb.ingress.kubernetes.io/scheme: internet-facing
    alb.ingress.kubernetes.io/tags: Environment=dev,Team=test
    alb.ingress.kubernetes.io/target-type: ip
    alb.ingress.kubernetes.io/target-group-attributes: stickiness.enabled=true,stickiness.lb_cookie.duration_seconds=604800,stickiness.type=lb_cookie
    alb.ingress.kubernetes.io/load-balancer-attributes: idle_timeout.timeout_seconds=600
    alb.ingress.kubernetes.io/group.name: windmill
    alb.ingress.kubernetes.io/group.order: '10'
    alb.ingress.kubernetes.io/certificate-arn: my-certificatearn

GCP + GCE LB + managed certificates

windmill:
  baseDomain: "windmill.example.com"
  app:
    service:
      annotations:
        cloud.google.com/backend-config: '{"default": "session-config"}'
    smtpService:
      enabled: true
      annotations:
        cloud.google.com/l4-rbs: "enabled"
        # # for static ip (more info on https://cloud.google.com/kubernetes-engine/docs/concepts/service-load-balancer-parameters#spd-static-ip-parameters):
        # networking.gke.io/load-balancer-ip-addresses: <REGIONAL_IP_NAME>
  lsp:
    service:
      annotations:
        cloud.google.com/backend-config: '{"default": "session-config"}'
        
ingress:
  annotations:
    kubernetes.io/ingress.class: "gce"
    kubernetes.io/ingress.global-static-ip-name: <GLOBAL_IP_NAME>
    networking.gke.io/managed-certificates: managed-cert

Replace <GLOBAL_IP_NAME> with the name of a global static IP address you've created in GCP.

In addition to the above, you will need to apply the following resources for session affinity and managed certificates:

apiVersion: cloud.google.com/v1
kind: BackendConfig
metadata:
  name: windmill-backendconfig
spec:
  sessionAffinity:
    affinityType: "GENERATED_COOKIE"
    affinityCookieTtlSec: 86400 # max

apiVersion: networking.gke.io/v1
kind: ManagedCertificate
metadata:
  name: managed-cert
spec:
  domains:
    - windmill.example.com

Azure + app routing + keyvault certificate

windmill:
  baseDomain: "windmill.example.com"
  app:
    smtpService:
      enabled: true
      # # for static ip (more info on https://learn.microsoft.com/en-us/azure/aks/static-ip):
      # annotations:
      #   service.beta.kubernetes.io/azure-pip-name: <myAKSPublicIP>
ingress:
  annotations:
    kubernetes.azure.com/tls-cert-keyvault-uri: <KeyVaultCertificateUri>
  className: webapprouting.kubernetes.azure.com
  tls:
    - hosts:
      - "windmill.example.com"
      secretName: keyvault-windmill

You can find more details about SSL certificates with webapprouting in Azure here.

NGINX ingress + cert-manager:

windmill:
  baseDomain: "windmill.example.com"
  ...

ingress:
  className: "nginx"
  tls:
    - hosts:
        - "windmill.example.com"
      secretName: windmill-tls-cert
  annotations:
    cert-manager.io/issuer: "letsencrypt-prod"
    nginx.ingress.kubernetes.io/affinity: "cookie"
    nginx.ingress.kubernetes.io/affinity-mode: "persistent"
    nginx.ingress.kubernetes.io/session-cookie-name: "route"
...

You will also need to install cert-manager and configure an issuer. More details here and here. Cert-manager can also be used with the other cloud providers.

Generic

There are many ways to expose an app and it will depend on the requirements of your environment. If you don't want to use the included ingress and roll your own, you can just disable it. Overall, you want the following endpoints accessible included in the chart:

• windmill app on port 8000
• lsp application on port 3001
• metrics endpoints on port 8001 for the app and workers (ee only)
• windmill app smtp service on port 2525 for email triggers (need to be exposed on port 25)

If you are using Prometheus and if the enterprise edition is enabled, you can scrape the windmill-app-metrics service on port 8001 at /metrics endpoint to gather stats about the Windmill application.

Docker dind configruation

If you don't want to run Docker using the host's Docker engine, you can use Docker-in-Docker (dind). Below is the configuration:

windmill:
  workerGroups:
  - name: "native"
    replicas: 2
    volumes:
    - emptyDir: {}
      name: sock-dir
    - emptyDir: {}
      name: windmill-workspace
    volumeMounts:
    - mountPath: /var/run
      name: sock-dir
    - mountPath: /opt/windmill
      name: windmill-workspace
    extraContainers:
    - args:
      - --mtu=1450
      image: docker:27.2.1-dind
      imagePullPolicy: IfNotPresent
      name: dind
      resources: {}
      securityContext:
        privileged: true
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: File
      volumeMounts:
      - mountPath: /opt/windmill
        name: windmill-workspace
      - mountPath: /var/run
        name: sock-dir
    mode: "worker"

NOTE: the windmill-workspace volumeMount is used to share files between the dind container and the worker container.