在一個 Pod 中執行 MOF、Dash 和 MCP。
Linux amd64 · kubectl · 外部 PostgreSQL · 1 Pod · 3 容器
可信任內網 · 需 CREATEDB 權限 · 暫不支援強制 TLS 的 PostgreSQL。
映像檔來源
下載 mof.yaml
Docker Hub · MOF v2.0.15 · Dash v2.0.17 · MCP v2.0.3
下載 mof.yaml ↓檢視檔案
# MOF v2.0.15 + Dash v2.0.17 + MCP v2.0.3
# Edit database address, user and password before applying. Keep this file private.
# Keep the Service names mof, mof-mcp and dash, all in one namespace: the components find each
# other by them (mof:8080, mof-mcp:8001). Renamed Services can't connect.
# Private-network PostgreSQL without TLS. TLS-only databases are not supported by this release.
apiVersion: v1
kind: Namespace
metadata:
name: mof
---
apiVersion: v1
kind: Secret
metadata:
name: mof-database
namespace: mof
type: Opaque
stringData:
password: "REPLACE_WITH_YOUR_DATABASE_PASSWORD"
---
apiVersion: v1
kind: ConfigMap
metadata:
name: mof-config
namespace: mof
data:
RK_POSTGRES_0_ADDR: "db.example.com:5432"
RK_POSTGRES_0_USER: "mof"
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof
namespace: mof
spec:
replicas: 1
# First-release template: stop the old Pod before starting the new version.
strategy:
type: Recreate
selector:
matchLabels:
app: mof
template:
metadata:
labels:
app: mof
# Prometheus discovery by annotation; see the monitoring guide for a PodMonitor
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "8080"
prometheus.io/path: /internal/v1/mof/metrics
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
terminationGracePeriodSeconds: 120
containers:
- name: mof
image: mofcloud/mof:v2.0.15
envFrom:
- configMapRef:
name: mof-config
env:
- name: RK_POSTGRES_0_PASS
valueFrom:
secretKeyRef:
name: mof-database
key: password
ports:
- name: api
containerPort: 8080
startupProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
failureThreshold: 60
readinessProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "500m"
memory: "1Gi"
volumeMounts:
- name: offer-cache
mountPath: /usr/src/mof/offer-cache
volumes:
- name: offer-cache
emptyDir: {}
---
apiVersion: v1
kind: Service
metadata:
name: mof
namespace: mof
spec:
selector:
app: mof
ports:
- name: api
port: 8080
targetPort: api
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof-mcp
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: mof-mcp
template:
metadata:
labels:
app: mof-mcp
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: mof-mcp
image: mofcloud/mof-mcp:v2.0.3
ports:
- name: mcp
containerPort: 8001
readinessProbe:
exec:
command: ["node", "/app/build/healthcheck.js"]
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "100m"
memory: "128Mi"
---
apiVersion: v1
kind: Service
metadata:
name: mof-mcp
namespace: mof
spec:
selector:
app: mof-mcp
ports:
- name: mcp
port: 8001
targetPort: mcp
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: dash
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: dash
template:
metadata:
labels:
app: dash
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: dash
image: mofcloud/dash:v2.0.17
ports:
- name: web
containerPort: 80
readinessProbe:
httpGet:
path: /
port: web
periodSeconds: 10
resources:
requests:
cpu: "100m"
memory: "64Mi"
---
apiVersion: v1
kind: Service
metadata:
name: dash
namespace: mof
spec:
selector:
app: dash
ports:
- name: web
port: 80
targetPort: web
type: ClusterIP
阿里雲 ACR · MOF v2.0.15 · Dash v2.0.17 · MCP v2.0.3
下載 mof.yaml ↓檢視檔案
# MOF v2.0.15 + Dash v2.0.17 + MCP v2.0.3
# Edit database address, user and password before applying. Keep this file private.
# Keep the Service names mof, mof-mcp and dash, all in one namespace: the components find each
# other by them (mof:8080, mof-mcp:8001). Renamed Services can't connect.
# Private-network PostgreSQL without TLS. TLS-only databases are not supported by this release.
apiVersion: v1
kind: Namespace
metadata:
name: mof
---
apiVersion: v1
kind: Secret
metadata:
name: mof-database
namespace: mof
type: Opaque
stringData:
password: "REPLACE_WITH_YOUR_DATABASE_PASSWORD"
---
apiVersion: v1
kind: ConfigMap
metadata:
name: mof-config
namespace: mof
data:
RK_POSTGRES_0_ADDR: "db.example.com:5432"
RK_POSTGRES_0_USER: "mof"
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof
namespace: mof
spec:
replicas: 1
# First-release template: stop the old Pod before starting the new version.
strategy:
type: Recreate
selector:
matchLabels:
app: mof
template:
metadata:
labels:
app: mof
# Prometheus discovery by annotation; see the monitoring guide for a PodMonitor
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "8080"
prometheus.io/path: /internal/v1/mof/metrics
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
terminationGracePeriodSeconds: 120
containers:
- name: mof
image: registry.cn-hangzhou.aliyuncs.com/mofcloud/mof:v2.0.15
envFrom:
- configMapRef:
name: mof-config
env:
- name: RK_POSTGRES_0_PASS
valueFrom:
secretKeyRef:
name: mof-database
key: password
ports:
- name: api
containerPort: 8080
startupProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
failureThreshold: 60
readinessProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "500m"
memory: "1Gi"
volumeMounts:
- name: offer-cache
mountPath: /usr/src/mof/offer-cache
volumes:
- name: offer-cache
emptyDir: {}
---
apiVersion: v1
kind: Service
metadata:
name: mof
namespace: mof
spec:
selector:
app: mof
ports:
- name: api
port: 8080
targetPort: api
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof-mcp
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: mof-mcp
template:
metadata:
labels:
app: mof-mcp
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: mof-mcp
image: registry.cn-hangzhou.aliyuncs.com/mofcloud/mof-mcp:v2.0.3
ports:
- name: mcp
containerPort: 8001
readinessProbe:
exec:
command: ["node", "/app/build/healthcheck.js"]
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "100m"
memory: "128Mi"
---
apiVersion: v1
kind: Service
metadata:
name: mof-mcp
namespace: mof
spec:
selector:
app: mof-mcp
ports:
- name: mcp
port: 8001
targetPort: mcp
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: dash
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: dash
template:
metadata:
labels:
app: dash
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: dash
image: registry.cn-hangzhou.aliyuncs.com/mofcloud/dash:v2.0.17
ports:
- name: web
containerPort: 80
readinessProbe:
httpGet:
path: /
port: web
periodSeconds: 10
resources:
requests:
cpu: "100m"
memory: "64Mi"
---
apiVersion: v1
kind: Service
metadata:
name: dash
namespace: mof
spec:
selector:
app: dash
ports:
- name: web
port: 80
targetPort: web
type: ClusterIP
GHCR · MOF v2.0.15 · Dash v2.0.17 · MCP v2.0.3
下載 mof.yaml ↓檢視檔案
# MOF v2.0.15 + Dash v2.0.17 + MCP v2.0.3
# Edit database address, user and password before applying. Keep this file private.
# Keep the Service names mof, mof-mcp and dash, all in one namespace: the components find each
# other by them (mof:8080, mof-mcp:8001). Renamed Services can't connect.
# Private-network PostgreSQL without TLS. TLS-only databases are not supported by this release.
apiVersion: v1
kind: Namespace
metadata:
name: mof
---
apiVersion: v1
kind: Secret
metadata:
name: mof-database
namespace: mof
type: Opaque
stringData:
password: "REPLACE_WITH_YOUR_DATABASE_PASSWORD"
---
apiVersion: v1
kind: ConfigMap
metadata:
name: mof-config
namespace: mof
data:
RK_POSTGRES_0_ADDR: "db.example.com:5432"
RK_POSTGRES_0_USER: "mof"
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof
namespace: mof
spec:
replicas: 1
# First-release template: stop the old Pod before starting the new version.
strategy:
type: Recreate
selector:
matchLabels:
app: mof
template:
metadata:
labels:
app: mof
# Prometheus discovery by annotation; see the monitoring guide for a PodMonitor
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "8080"
prometheus.io/path: /internal/v1/mof/metrics
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
terminationGracePeriodSeconds: 120
containers:
- name: mof
image: ghcr.io/mofcloud/mof:v2.0.15
envFrom:
- configMapRef:
name: mof-config
env:
- name: RK_POSTGRES_0_PASS
valueFrom:
secretKeyRef:
name: mof-database
key: password
ports:
- name: api
containerPort: 8080
startupProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
failureThreshold: 60
readinessProbe:
httpGet:
path: /v1/mof/site-info
port: api
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "500m"
memory: "1Gi"
volumeMounts:
- name: offer-cache
mountPath: /usr/src/mof/offer-cache
volumes:
- name: offer-cache
emptyDir: {}
---
apiVersion: v1
kind: Service
metadata:
name: mof
namespace: mof
spec:
selector:
app: mof
ports:
- name: api
port: 8080
targetPort: api
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: mof-mcp
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: mof-mcp
template:
metadata:
labels:
app: mof-mcp
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: mof-mcp
image: ghcr.io/mofcloud/mof-mcp:v2.0.3
ports:
- name: mcp
containerPort: 8001
readinessProbe:
exec:
command: ["node", "/app/build/healthcheck.js"]
periodSeconds: 10
timeoutSeconds: 5
resources:
requests:
cpu: "100m"
memory: "128Mi"
---
apiVersion: v1
kind: Service
metadata:
name: mof-mcp
namespace: mof
spec:
selector:
app: mof-mcp
ports:
- name: mcp
port: 8001
targetPort: mcp
type: ClusterIP
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: dash
namespace: mof
spec:
replicas: 1
selector:
matchLabels:
app: dash
template:
metadata:
labels:
app: dash
spec:
enableServiceLinks: false
nodeSelector:
kubernetes.io/arch: amd64
containers:
- name: dash
image: ghcr.io/mofcloud/dash:v2.0.17
ports:
- name: web
containerPort: 80
readinessProbe:
httpGet:
path: /
port: web
periodSeconds: 10
resources:
requests:
cpu: "100m"
memory: "64Mi"
---
apiVersion: v1
kind: Service
metadata:
name: dash
namespace: mof
spec:
selector:
app: dash
ports:
- name: web
port: 80
targetPort: web
type: ClusterIP
不要修改 Service 名稱 mof、mof-mcp 和 dash,三者放在同一個命名空間。元件之間依這些名稱互相存取(mof:8080、mof-mcp:8001),因此不需其他設定;改了名稱就連不上。
資料庫連線
在 mof.yaml 裡填寫既有 PostgreSQL 的連線資訊:
RK_POSTGRES_0_ADDR- 位址:連接埠
RK_POSTGRES_0_USER- 使用者名稱
stringData.password- 資料庫既有密碼
密碼保留引號,檔案不要公開。
部署
kubectl apply -f mof.yaml
kubectl -n mof rollout status deployment/mof --timeout=10m
kubectl -n mof rollout status deployment/mof-mcp deployment/dash三個都就緒後,查看 Admin 密碼:
kubectl -n mof logs deployment/mof | grep "Admin password"開啟 MOF
kubectl -n mof port-forward service/dash 8000:80保持這個終端機執行。
http://127.0.0.1:8000/admin設定授權並建立使用者。
http://127.0.0.1:8000使用 MOF 帳號登入。
進階設定
PostgreSQL TLS 相容性
MOF 的 TLS 設定僅涵蓋 Mof 和 AsyncJobMof,未涵蓋雲端廠商資料庫。此範本僅適用於可信任內網,無法連線強制 TLS 的 PostgreSQL。
網域、HTTPS 與選用 OAuth
使用既有的 Ingress 控制器並將網域指向入口,填寫 class、網域和 mof 命名空間中既有的 TLS Secret。
下載 ingress.yaml ↓# Requires an installed Ingress controller and an existing TLS Secret.
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
name: mof
namespace: mof
spec:
ingressClassName: nginx
tls:
- hosts: [mof.example.com]
secretName: mof-web-tls
rules:
- host: mof.example.com
http:
paths:
- path: /
pathType: Prefix
backend:
service:
name: dash
port:
number: 80
kubectl apply -f ingress.yaml
kubectl -n mof get ingress在 Admin → 設定 → MCP 中啟用存取,路徑為 /v1/mof/remote-mcp。權杖不需要 OAuth;瀏覽器 OAuth 需要 HTTPS、MOF_PUBLIC_URL 和 MOF_MCP_OAUTH_SECRET。
健康檢查與疑難排解
kubectl -n mof get pods
kubectl -n mof logs deployment/mof --tail=100
kubectl -n mof logs deployment/mof-mcp --tail=60
kubectl -n mof logs deployment/dash --tail=60
kubectl -n mof exec deployment/mof -- wget -q -O - http://mof-mcp:8001/health預期:mof、mof-mcp、dash 都是 Running 且 Ready,MCP ready: true。啟動失敗時請檢查映像檔拉取、節點架構和資料庫連線。
範本參數說明
RK_POSTGRES_0_ADDR外部 PostgreSQL 位址:連接埠,需能從 Pod 存取。RK_POSTGRES_0_USER既有的資料庫使用者,需能建立 MOF 及雲端廠商資料庫。mof-database / password存放在 Secret 中的資料庫密碼,僅注入 MOF。Services mof / mof-mcp / dash各一個 Deployment 和 Service,依 Service 名稱互相存取,不需設定。replicas / strategy1 / Recreate。升級會短暫中斷,本範本不提供多副本高可用性。resources.requests排程預留量,並非記憶體上限。請依帳單規模調整 MOF 資源並監控實際用量。offer-cache / emptyDir可重建的公開價格快取,Pod 重建後重新載入;帳單資料存於 PostgreSQL。ingressClassName / host / tls對應已安裝的控制器、存取網域和 TLS Secret;本機連接埠轉送驗證時不需要。更新、資料與擴充
先備份 PostgreSQL,再修改映像檔版本並套用 mof.yaml。修改 ConfigMap 或 Secret 後需重新啟動;Recreate 升級會中斷服務。
kubectl -n mof rollout restart deployment/mof
kubectl -n mof rollout status deployment/mof --timeout=10m本範本僅支援單一副本;隨需 Kubernetes Job 尚未實作。應用程式資料保存在外部 PostgreSQL,價格快取可重建。