[Share] Configuration Kubernetes (k8s) for one-api

[H0llyW00dzZ]

&

---

Introduction:
I am sharing this Kubernetes (k8s) configuration for one-api because it has proven to be stable when used with this repository.

• The backend is written in Go and uses the Gin package

Note

The backend, written in Go and using the Gin package, is a microservice architecture. Therefore, you don't need to create many pods for one-api. A single pod, along with hpa, is sufficient.

---

Deployment:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: one-api
spec:
  replicas: 3
  selector:
    matchLabels:
      app: one-api
  template:
    metadata:
      labels:
        app: one-api
    spec:
      containers:
      - name: one-api
        image: ghcr.io/songquanpeng/one-api:v0.6.0-alpha.3 # Modify this as well
        ports:
        - containerPort: 3000
        resources:
          requests:
            memory: "256Mi"
            cpu: "250m"
          limits:
            memory: "512Mi"
            cpu: "500m"
        volumeMounts:
        - mountPath: "/data"
          name: data-volume
        env:
        - name: SESSION_SECRET
          value: yZlNPz2gciO8HqVCrhn0 # this just random string
        - name: SYNC_FREQUENCY
          value: "60"
        - name: CHANNEL_UPDATE_FREQUENCY
          value: "1440"
        - name: CHANNEL_TEST_FREQUENCY
          value: "1440"
        - name: POLLING_INTERVAL
          value: "5"
        - name: NODE_TYPE
          value: master
        - name: SQL_DSN
          valueFrom: # this important you need to use secrets, for example I am using mysql for this
            secretKeyRef:
              key: sql-dsn
              name: my-db-secret
      volumes:
      - name: data-volume
        persistentVolumeClaim:
          claimName: one-api-storage

Note

For the deployment, even if you initially set replica 3, it will be automated adjusted to 1 after you set up the Horizontal Pod Autoscaler (HPA), as shown below.

Horizontal Pod Autoscaler (HPA):

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: one-api-hpa
  namespace: one-api-namespace # this namespace
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: one-api
  minReplicas: 1
  maxReplicas: 5
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 80
    - type: Resource
      resource:
        name: memory
        target:
          type: Utilization
          averageUtilization: 80

Storage:

apiVersion: v1
kind: PersistentVolume
metadata:
  name: one-api-storage
spec:
  capacity:
    storage: 5Gi
  accessModes:
    - ReadWriteOnce
  persistentVolumeReclaimPolicy: Retain
  storageClassName: standard
  hostPath:
    path: /mnt/data

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: one-api-storage
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: "standard"
  resources:
    requests:
      storage: 5Gi

Note

This storage can be used for SQLite or any other storage needs you may have.

---

Note

The configuration below is for nginx used with one-api. You won't need to configure it manually as you would with docker.

---

Services:

apiVersion: v1
kind: Service
metadata:
  name: one-api-service
spec:
  selector:
    app: one-api
  ports:
    - protocol: TCP
      port: 80
      targetPort: 3000

Cluster Role:

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: ingress-nginx-leader-election-rolebinding
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: ClusterRole
  name: ingress-nginx-leader-election-role
subjects:
- kind: ServiceAccount
  name: ingress-nginx
  namespace: ingress-nginx

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: ingress-nginx
rules:
  - apiGroups: [""]
    resources: ["configmaps", "endpoints", "pods", "services", "events", "secrets"] # Adding Secrets for CertManager
    verbs: ["get", "list", "watch"]
  - apiGroups: ["extensions", "networking.k8s.io"]
    resources: ["ingresses", "ingressclasses", "ingresses/status"]
    verbs: ["get", "list", "watch", "update"]
  - apiGroups: ["discovery.k8s.io"]
    resources: ["endpointslices"]
    verbs: ["get", "list", "watch"]

Ingress Nginx

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: one-api-ingress
  namespace: one-api-namespace
  annotations:
    kubernetes.io/ingress.class: "nginx" # This may vary depending on your Ingress controller
    cert-manager.io/cluster-issuer: "letsencrypt-prod" # Ensure this matches the name of your ClusterIssuer
spec:
  ingressClassName: nginx
  rules:
  - host: your-site.com # modify this
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service:
            name: one-api-service
            port:
              number: 80
  tls: # Define the TLS section to enable HTTPS
  - hosts:
    - your-site.com # modify this
    secretName: your-site-com-tls # Cert-Manager will store the acquired certificate here

[H0llyW00dzZ]

Another Note

Note

For mysql you need vpa configuration, might I will share it later

[H0llyW00dzZ]

Show off it's so cheap and stable

(base) root@H0llyW00dzZ:/# kubectl top pods
NAME                       CPU(cores)   MEMORY(bytes)
mysql-ochjeq9dpo-1k14      5m           403Mi
one-api-0tb4urx12y-0fzj2   0m           44Mi

[H0llyW00dzZ]

Another show off still can handle 15 million tokens without stupid memory leaks that I don't fucking understand

[H0llyW00dzZ]

bump docs updated

[H0llyW00dzZ]

bump zero downtime

$ kubectl get pods
NAME                       READY   STATUS    RESTARTS    AGE
mysql-krq66e7q1m-6h9qd     1/1     Running    0          3d15h
one-api-p2caldes54-7xzgn   1/1     Running    0          3d15h

[H0llyW00dzZ]

Below is proof of the successful configuration plus stable with better performance:

(base) root@H0llyW00dzZ:/# kubectl get hpa
NAME          REFERENCE            TARGETS           MINPODS   MAXPODS   REPLICAS   AGE
one-api-hpa   Deployment/one-api   0%/80%, 16%/80%   1         10        1          6d8h