Merge pull request #114 from laszlocph/deploy-to-k8s

K8s DinD deployment
This commit is contained in:
Laszlo Fogas 2020-04-28 21:43:42 +02:00 committed by GitHub
commit 6ee3adc72c
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 166 additions and 1 deletions

View file

@ -130,6 +130,7 @@ services:
ports: ["5432"]
environment:
- POSTGRES_USER=postgres
- POSTGRES_HOST_AUTH_METHOD=trust
mysql:
image: mysql:5.6.27
ports: ["3306"]

View file

@ -190,3 +190,167 @@ A [Prometheus endpoint](/administration/prometheus) is exposed.
## Behind a proxy
See the [proxy guide](/administration/proxy) if you want to see a setup behind Apache, Nginx, Caddy or ngrok.
## Deploying on Kubernetes
Woodpecker does not support Kubernetes natively, but being a container first CI engine, it can be deployed to Kubernetes.
The following yamls represent a server (backed by sqlite and Persistent Volumes) and an agent deployment. The agents can be scaled by the `replica` field.
By design, Woodpecker spins up a new container for each workflow step. It talks to the Docker agent to do that.
However in Kubernetes, the Docker agent is not accessible, therefore this deployment follows a Docker in Docker setup and we deploy a DinD sidecar with the agent.
Build step containers are started up within the agent pod.
Warning: this approach requires `privileged` access. Also DinD's reputation hasn't been too high in the early days of Docker - this changed somewhat over time, and there are organizations succeeding with this approach.
server.yaml
```yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: woodpecker
namespace: tools
labels:
app: woodpecker
spec:
replicas: 1
selector:
matchLabels:
app: woodpecker
template:
metadata:
labels:
app: woodpecker
annotations:
prometheus.io/scrape: 'true'
spec:
containers:
- image: laszlocloud/woodpecker-server:v0.9.2
imagePullPolicy: Always
name: woodpecker
env:
- name: "DRONE_ADMIN"
value: "xxx"
- name: "DRONE_HOST"
value: "https://xxx"
- name: "DRONE_GITHUB"
value: "true"
- name: "DRONE_GITHUB_CLIENT"
value: "xxx"
- name: "DRONE_GITHUB_SECRET"
value: "xxx"
- name: "DRONE_SECRET"
value: "xxx"
volumeMounts:
- name: sqlite-volume
mountPath: /var/lib/drone
volumes:
- name: sqlite-volume
persistentVolumeClaim:
claimName: woodpecker-pvc
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: woodpecker-pvc
namespace: tools
spec:
accessModes:
- ReadWriteOnce
storageClassName: local-path
resources:
requests:
storage: 10Gi
---
kind: Service
apiVersion: v1
metadata:
name: woodpecker
namespace: tools
spec:
type: ClusterIP
selector:
app: woodpecker
ports:
- protocol: TCP
name: http
port: 80
targetPort: 8000
- protocol: TCP
name: grpc
port: 9000
targetPort: 9000
---
kind: Ingress
apiVersion: extensions/v1beta1
metadata:
name: woodpecker
namespace: tools
spec:
tls:
- hosts:
- xxx
secretName: xxx
rules:
- host: xxx
http:
paths:
- backend:
serviceName: woodpecker
servicePort: 80
```
agent.yaml
```yaml
apiVersion: apps/v1
kind: Deployment
metadata:
name: woodpecker-agent
namespace: tools
labels:
app: woodpecker-agent
spec:
selector:
matchLabels:
app: woodpecker-agent
replicas: 2
template:
metadata:
annotations:
labels:
app: woodpecker-agent
spec:
containers:
- name: agent
image: laszlocloud/woodpecker-agent:v0.9.2
imagePullPolicy: Always
ports:
- name: http
containerPort: 3000
protocol: TCP
env:
- name: DRONE_SERVER
value: woodpecker.tools.svc.cluster.local:9000
- name: DRONE_SECRET
value: "xxx"
- name: DOCKER_HOST
value: tcp://localhost:2375
resources:
limits:
cpu: 2
memory: 2Gi
- name: dind
image: "docker:19.03.5-dind"
env:
- name: DOCKER_DRIVER
value: overlay2
- name: DOCKER_TLS_CERTDIR
value: "" # due to https://github.com/docker-library/docker/pull/166 & https://gitlab.com/gitlab-org/gitlab-runner/issues/4512
resources:
limits:
cpu: 1
memory: 2Gi
securityContext:
privileged: true
```