首次提交：初始化项目
This commit is contained in:
fei
380
009-基础设施/007-keda/scalers/KEDA-自动缩容到0-配置指南.md
Normal file
380
009-基础设施/007-keda/scalers/KEDA-自动缩容到0-配置指南.md
Normal file
@@ -0,0 +1,380 @@
|
||||
# KEDA HTTP Add-on 自动缩容到 0 配置指南
|
||||
|
||||
本指南说明如何使用 KEDA HTTP Add-on 实现应用在无流量时自动缩容到 0，有访问时自动启动。
|
||||
|
||||
## 前提条件
|
||||
|
||||
1. K3s 集群已安装
|
||||
2. KEDA 已安装
|
||||
3. KEDA HTTP Add-on 已安装
|
||||
4. Traefik 作为 Ingress Controller
|
||||
|
||||
### 检查 KEDA HTTP Add-on 是否已安装
|
||||
|
||||
```bash
|
||||
kubectl get pods -n keda | grep http
|
||||
```
|
||||
|
||||
应该看到类似输出：
|
||||
```
|
||||
keda-add-ons-http-controller-manager-xxx 1/1 Running
|
||||
keda-add-ons-http-external-scaler-xxx 1/1 Running
|
||||
keda-add-ons-http-interceptor-xxx 1/1 Running
|
||||
```
|
||||
|
||||
### 如果未安装，执行以下命令安装
|
||||
|
||||
```bash
|
||||
helm repo add kedacore https://kedacore.github.io/charts
|
||||
helm repo update
|
||||
helm install http-add-on kedacore/keda-add-ons-http --namespace keda
|
||||
```
|
||||
|
||||
## 配置步骤
|
||||
|
||||
### 1. 准备应用的基础资源
|
||||
|
||||
确保你的应用已经有以下资源：
|
||||
- Deployment
|
||||
- Service
|
||||
- Namespace
|
||||
|
||||
示例：
|
||||
```yaml
|
||||
apiVersion: v1
|
||||
kind: Namespace
|
||||
metadata:
|
||||
name: myapp
|
||||
|
||||
---
|
||||
apiVersion: apps/v1
|
||||
kind: Deployment
|
||||
metadata:
|
||||
name: myapp
|
||||
namespace: myapp
|
||||
spec:
|
||||
replicas: 1
|
||||
selector:
|
||||
matchLabels:
|
||||
app: myapp
|
||||
template:
|
||||
metadata:
|
||||
labels:
|
||||
app: myapp
|
||||
spec:
|
||||
containers:
|
||||
- name: myapp
|
||||
image: your-image:tag
|
||||
ports:
|
||||
- containerPort: 80
|
||||
|
||||
---
|
||||
apiVersion: v1
|
||||
kind: Service
|
||||
metadata:
|
||||
name: myapp
|
||||
namespace: myapp
|
||||
spec:
|
||||
selector:
|
||||
app: myapp
|
||||
ports:
|
||||
- port: 80
|
||||
targetPort: 80
|
||||
```
|
||||
|
||||
### 2. 创建 HTTPScaledObject
|
||||
|
||||
这是实现自动缩容到 0 的核心配置。
|
||||
|
||||
```yaml
|
||||
apiVersion: http.keda.sh/v1alpha1
|
||||
kind: HTTPScaledObject
|
||||
metadata:
|
||||
name: myapp-http-scaler
|
||||
namespace: myapp # 必须与应用在同一个 namespace
|
||||
spec:
|
||||
hosts:
|
||||
- myapp.example.com # 你的域名
|
||||
pathPrefixes:
|
||||
- / # 匹配的路径前缀
|
||||
scaleTargetRef:
|
||||
name: myapp # Deployment 名称
|
||||
kind: Deployment
|
||||
apiVersion: apps/v1
|
||||
service: myapp # Service 名称
|
||||
port: 80 # Service 端口
|
||||
replicas:
|
||||
min: 0 # 空闲时缩容到 0
|
||||
max: 10 # 最多扩容到 10 个副本
|
||||
scalingMetric:
|
||||
requestRate:
|
||||
granularity: 1s
|
||||
targetValue: 100 # 每秒 100 个请求时扩容
|
||||
window: 1m
|
||||
scaledownPeriod: 300 # 5 分钟（300秒）无流量后缩容到 0
|
||||
```
|
||||
|
||||
**重要参数说明：**
|
||||
- `hosts`: 你的应用域名
|
||||
- `scaleTargetRef.name`: 你的 Deployment 名称
|
||||
- `scaleTargetRef.service`: 你的 Service 名称
|
||||
- `scaleTargetRef.port`: 你的 Service 端口
|
||||
- `replicas.min: 0`: 允许缩容到 0
|
||||
- `scaledownPeriod`: 无流量后多久缩容（秒）
|
||||
|
||||
### 3. 创建 Traefik IngressRoute
|
||||
|
||||
**重要：IngressRoute 必须在 keda namespace 中创建**，因为它需要引用 keda namespace 的拦截器服务。
|
||||
|
||||
```yaml
|
||||
apiVersion: traefik.io/v1alpha1
|
||||
kind: IngressRoute
|
||||
metadata:
|
||||
name: myapp-ingress
|
||||
namespace: keda # 注意：必须在 keda namespace
|
||||
spec:
|
||||
entryPoints:
|
||||
- web # HTTP 入口
|
||||
# - websecure # 如果需要 HTTPS，添加这个
|
||||
routes:
|
||||
- match: Host(`myapp.example.com`) # 你的域名
|
||||
kind: Rule
|
||||
services:
|
||||
- name: keda-add-ons-http-interceptor-proxy
|
||||
port: 8080
|
||||
```
|
||||
|
||||
**如果需要 HTTPS，添加 TLS 配置：**
|
||||
|
||||
```yaml
|
||||
apiVersion: traefik.io/v1alpha1
|
||||
kind: IngressRoute
|
||||
metadata:
|
||||
name: myapp-ingress
|
||||
namespace: keda
|
||||
spec:
|
||||
entryPoints:
|
||||
- websecure
|
||||
routes:
|
||||
- match: Host(`myapp.example.com`)
|
||||
kind: Rule
|
||||
services:
|
||||
- name: keda-add-ons-http-interceptor-proxy
|
||||
port: 8080
|
||||
tls:
|
||||
certResolver: letsencrypt # 你的证书解析器
|
||||
```
|
||||
|
||||
### 4. 完整配置文件模板
|
||||
|
||||
将以下内容保存为 `myapp-keda-scaler.yaml`，并根据你的应用修改相应的值：
|
||||
|
||||
```yaml
|
||||
---
|
||||
# HTTPScaledObject - 实现自动缩容到 0
|
||||
apiVersion: http.keda.sh/v1alpha1
|
||||
kind: HTTPScaledObject
|
||||
metadata:
|
||||
name: myapp-http-scaler
|
||||
namespace: myapp # 改为你的 namespace
|
||||
spec:
|
||||
hosts:
|
||||
- myapp.example.com # 改为你的域名
|
||||
pathPrefixes:
|
||||
- /
|
||||
scaleTargetRef:
|
||||
name: myapp # 改为你的 Deployment 名称
|
||||
kind: Deployment
|
||||
apiVersion: apps/v1
|
||||
service: myapp # 改为你的 Service 名称
|
||||
port: 80 # 改为你的 Service 端口
|
||||
replicas:
|
||||
min: 0
|
||||
max: 10
|
||||
scalingMetric:
|
||||
requestRate:
|
||||
granularity: 1s
|
||||
targetValue: 100
|
||||
window: 1m
|
||||
scaledownPeriod: 300 # 5 分钟无流量后缩容
|
||||
|
||||
---
|
||||
# Traefik IngressRoute - 路由流量到 KEDA 拦截器
|
||||
apiVersion: traefik.io/v1alpha1
|
||||
kind: IngressRoute
|
||||
metadata:
|
||||
name: myapp-ingress
|
||||
namespace: keda # 必须在 keda namespace
|
||||
spec:
|
||||
entryPoints:
|
||||
- web
|
||||
routes:
|
||||
- match: Host(`myapp.example.com`) # 改为你的域名
|
||||
kind: Rule
|
||||
services:
|
||||
- name: keda-add-ons-http-interceptor-proxy
|
||||
port: 8080
|
||||
```
|
||||
|
||||
### 5. 应用配置
|
||||
|
||||
```bash
|
||||
kubectl apply -f myapp-keda-scaler.yaml
|
||||
```
|
||||
|
||||
### 6. 验证配置
|
||||
|
||||
```bash
|
||||
# 查看 HTTPScaledObject 状态
|
||||
kubectl get httpscaledobject -n myapp
|
||||
|
||||
# 应该看到 READY = True
|
||||
# NAME TARGETWORKLOAD TARGETSERVICE MINREPLICAS MAXREPLICAS AGE READY
|
||||
# myapp-http-scaler apps/v1/Deployment/myapp myapp:80 0 10 10s True
|
||||
|
||||
# 查看 IngressRoute
|
||||
kubectl get ingressroute -n keda
|
||||
|
||||
# 查看当前 Pod 数量
|
||||
kubectl get pods -n myapp
|
||||
```
|
||||
|
||||
## 工作原理
|
||||
|
||||
1. **有流量时**：
|
||||
- 用户访问 `myapp.example.com`
|
||||
- Traefik 将流量路由到 KEDA HTTP 拦截器
|
||||
- 拦截器检测到请求，通知 KEDA 启动 Pod
|
||||
- Pod 启动后（5-10秒），拦截器将流量转发到应用
|
||||
- 用户看到正常响应（首次访问可能有延迟）
|
||||
|
||||
2. **无流量时**：
|
||||
- 5 分钟（scaledownPeriod）无请求后
|
||||
- KEDA 自动将 Deployment 缩容到 0
|
||||
- 不消耗任何计算资源
|
||||
|
||||
## 常见问题排查
|
||||
|
||||
### 1. 访问返回 404
|
||||
|
||||
**检查 IngressRoute 是否在 keda namespace：**
|
||||
```bash
|
||||
kubectl get ingressroute -n keda | grep myapp
|
||||
```
|
||||
|
||||
如果不在，删除并重新创建：
|
||||
```bash
|
||||
kubectl delete ingressroute myapp-ingress -n myapp # 删除错误的
|
||||
kubectl apply -f myapp-keda-scaler.yaml # 重新创建
|
||||
```
|
||||
|
||||
### 2. HTTPScaledObject READY = False
|
||||
|
||||
**查看详细错误信息：**
|
||||
```bash
|
||||
kubectl describe httpscaledobject myapp-http-scaler -n myapp
|
||||
```
|
||||
|
||||
**常见错误：**
|
||||
- `workload already managed by ScaledObject`: 删除旧的 ScaledObject
|
||||
```bash
|
||||
kubectl delete scaledobject myapp-scaler -n myapp
|
||||
```
|
||||
|
||||
### 3. Pod 没有自动缩容到 0
|
||||
|
||||
**检查是否有旧的 ScaledObject 阻止缩容：**
|
||||
```bash
|
||||
kubectl get scaledobject -n myapp
|
||||
```
|
||||
|
||||
如果有，删除它：
|
||||
```bash
|
||||
kubectl delete scaledobject <name> -n myapp
|
||||
```
|
||||
|
||||
### 4. 查看 KEDA 拦截器日志
|
||||
|
||||
```bash
|
||||
kubectl logs -n keda -l app.kubernetes.io/name=keda-add-ons-http-interceptor --tail=50
|
||||
```
|
||||
|
||||
### 5. 测试拦截器是否工作
|
||||
|
||||
```bash
|
||||
# 获取拦截器服务 IP
|
||||
kubectl get svc keda-add-ons-http-interceptor-proxy -n keda
|
||||
|
||||
# 直接测试拦截器
|
||||
curl -H "Host: myapp.example.com" http://<CLUSTER-IP>:8080
|
||||
```
|
||||
|
||||
## 调优建议
|
||||
|
||||
### 调整缩容时间
|
||||
|
||||
根据你的应用特点调整 `scaledownPeriod`：
|
||||
|
||||
- **频繁访问的应用**：设置较长时间（如 600 秒 = 10 分钟）
|
||||
- **偶尔访问的应用**：设置较短时间（如 180 秒 = 3 分钟）
|
||||
- **演示/测试环境**：可以设置很短（如 60 秒 = 1 分钟）
|
||||
|
||||
```yaml
|
||||
scaledownPeriod: 600 # 10 分钟
|
||||
```
|
||||
|
||||
### 调整扩容阈值
|
||||
|
||||
根据应用负载调整 `targetValue`：
|
||||
|
||||
```yaml
|
||||
scalingMetric:
|
||||
requestRate:
|
||||
targetValue: 50 # 每秒 50 个请求时扩容（更敏感）
|
||||
```
|
||||
|
||||
### 调整最大副本数
|
||||
|
||||
```yaml
|
||||
replicas:
|
||||
min: 0
|
||||
max: 20 # 根据你的资源和需求调整
|
||||
```
|
||||
|
||||
## 监控和观察
|
||||
|
||||
### 实时监控 Pod 变化
|
||||
|
||||
```bash
|
||||
watch -n 2 'kubectl get pods -n myapp'
|
||||
```
|
||||
|
||||
### 查看 HTTPScaledObject 事件
|
||||
|
||||
```bash
|
||||
kubectl describe httpscaledobject myapp-http-scaler -n myapp
|
||||
```
|
||||
|
||||
### 查看 Deployment 副本数变化
|
||||
|
||||
```bash
|
||||
kubectl get deployment myapp -n myapp -w
|
||||
```
|
||||
|
||||
## 完整示例：navigation 应用
|
||||
|
||||
参考 `navigation-complete.yaml` 文件，这是一个完整的工作示例。
|
||||
|
||||
## 注意事项
|
||||
|
||||
1. **首次访问延迟**：Pod 从 0 启动需要 5-10 秒，用户首次访问会有延迟
|
||||
2. **数据库连接**：确保应用能够快速重新建立数据库连接
|
||||
3. **会话状态**：不要在 Pod 中存储会话状态，使用 Redis 等外部存储
|
||||
4. **健康检查**：配置合理的 readinessProbe，确保 Pod 就绪后才接收流量
|
||||
5. **资源限制**：设置合理的 resources limits，避免启动过慢
|
||||
|
||||
## 参考资源
|
||||
|
||||
- KEDA 官方文档: https://keda.sh/
|
||||
- KEDA HTTP Add-on: https://github.com/kedacore/http-add-on
|
||||
- Traefik IngressRoute: https://doc.traefik.io/traefik/routing/providers/kubernetes-crd/
|
||||
45
009-基础设施/007-keda/scalers/navigation-complete.yaml
Normal file
45
009-基础设施/007-keda/scalers/navigation-complete.yaml
Normal file
@@ -0,0 +1,45 @@
|
||||
---
|
||||
# HTTPScaledObject - 用于实现缩容到 0 的核心配置
|
||||
apiVersion: http.keda.sh/v1alpha1
|
||||
kind: HTTPScaledObject
|
||||
metadata:
|
||||
name: navigation-http-scaler
|
||||
namespace: navigation
|
||||
spec:
|
||||
hosts:
|
||||
- dh.u6.net3w.com
|
||||
pathPrefixes:
|
||||
- /
|
||||
scaleTargetRef:
|
||||
name: navigation
|
||||
kind: Deployment
|
||||
apiVersion: apps/v1
|
||||
service: navigation
|
||||
port: 80
|
||||
replicas:
|
||||
min: 0 # 空闲时缩容到 0
|
||||
max: 10 # 最多 10 个副本
|
||||
scalingMetric:
|
||||
requestRate:
|
||||
granularity: 1s
|
||||
targetValue: 100 # 每秒 100 个请求时扩容
|
||||
window: 1m
|
||||
scaledownPeriod: 300 # 5 分钟无流量后缩容到 0
|
||||
|
||||
---
|
||||
# Traefik IngressRoute - 将流量路由到 KEDA HTTP Add-on 的拦截器
|
||||
# 注意：必须在 keda namespace 中才能引用该 namespace 的服务
|
||||
apiVersion: traefik.io/v1alpha1
|
||||
kind: IngressRoute
|
||||
metadata:
|
||||
name: navigation-ingress
|
||||
namespace: keda
|
||||
spec:
|
||||
entryPoints:
|
||||
- web
|
||||
routes:
|
||||
- match: Host(`dh.u6.net3w.com`)
|
||||
kind: Rule
|
||||
services:
|
||||
- name: keda-add-ons-http-interceptor-proxy
|
||||
port: 8080
|
||||
24
009-基础设施/007-keda/scalers/navigation-http-scaler.yaml
Normal file
24
009-基础设施/007-keda/scalers/navigation-http-scaler.yaml
Normal file
@@ -0,0 +1,24 @@
|
||||
apiVersion: http.keda.sh/v1alpha1
|
||||
kind: HTTPScaledObject
|
||||
metadata:
|
||||
name: navigation-http-scaler
|
||||
namespace: navigation
|
||||
spec:
|
||||
hosts:
|
||||
- dh.u6.net3w.com
|
||||
pathPrefixes:
|
||||
- /
|
||||
scaleTargetRef:
|
||||
name: navigation
|
||||
kind: Deployment
|
||||
apiVersion: apps/v1
|
||||
service: navigation
|
||||
port: 80
|
||||
replicas:
|
||||
min: 0 # 空闲时缩容到 0
|
||||
max: 10 # 最多 10 个副本
|
||||
scalingMetric:
|
||||
requestRate:
|
||||
granularity: 1s
|
||||
targetValue: 100 # 每秒 100 个请求时扩容
|
||||
window: 1m
|
||||
19
009-基础设施/007-keda/scalers/navigation-ingress-http.yaml
Normal file
19
009-基础设施/007-keda/scalers/navigation-ingress-http.yaml
Normal file
@@ -0,0 +1,19 @@
|
||||
apiVersion: networking.k8s.io/v1
|
||||
kind: Ingress
|
||||
metadata:
|
||||
name: navigation-ingress
|
||||
namespace: navigation
|
||||
annotations:
|
||||
traefik.ingress.kubernetes.io/router.entrypoints: web
|
||||
spec:
|
||||
rules:
|
||||
- host: dh.u6.net3w.com
|
||||
http:
|
||||
paths:
|
||||
- path: /
|
||||
pathType: Prefix
|
||||
backend:
|
||||
service:
|
||||
name: keda-add-ons-http-interceptor-proxy
|
||||
port:
|
||||
number: 8080
|
||||
23
009-基础设施/007-keda/scalers/navigation-scaler.yaml
Normal file
23
009-基础设施/007-keda/scalers/navigation-scaler.yaml
Normal file
@@ -0,0 +1,23 @@
|
||||
apiVersion: keda.sh/v1alpha1
|
||||
kind: ScaledObject
|
||||
metadata:
|
||||
name: navigation-scaler
|
||||
namespace: navigation
|
||||
spec:
|
||||
scaleTargetRef:
|
||||
name: navigation
|
||||
minReplicaCount: 1 # 至少保持 1 个副本（HPA 限制）
|
||||
maxReplicaCount: 10 # 最多 10 个副本
|
||||
pollingInterval: 15 # 每 15 秒检查一次
|
||||
cooldownPeriod: 180 # 缩容冷却期 3 分钟
|
||||
triggers:
|
||||
- type: prometheus
|
||||
metadata:
|
||||
serverAddress: http://kube-prometheus-stack-prometheus.monitoring.svc.cluster.local:9090
|
||||
metricName: nginx_http_requests_total
|
||||
query: sum(rate(nginx_http_requests_total{namespace="navigation"}[1m]))
|
||||
threshold: "10" # 每分钟超过 10 个请求时启动
|
||||
- type: cpu
|
||||
metricType: Utilization
|
||||
metadata:
|
||||
value: "60" # CPU 使用率超过 60% 时扩容
|
||||
261
009-基础设施/007-keda/scalers/postgresql-说明.md
Normal file
261
009-基础设施/007-keda/scalers/postgresql-说明.md
Normal file
@@ -0,0 +1,261 @@
|
||||
# ⚠️ PostgreSQL 不适合使用 KEDA 自动扩缩容
|
||||
|
||||
## 问题说明
|
||||
|
||||
对于传统的 PostgreSQL 架构，直接通过 KEDA 增加副本数会导致：
|
||||
|
||||
### 1. 存储冲突
|
||||
- 多个 Pod 尝试挂载同一个 PVC
|
||||
- ReadWriteOnce 存储只能被一个 Pod 使用
|
||||
- 会导致 Pod 启动失败
|
||||
|
||||
### 2. 数据损坏风险
|
||||
- 如果使用 ReadWriteMany 存储，多个实例同时写入会导致数据损坏
|
||||
- PostgreSQL 不支持多主写入
|
||||
- 没有锁机制保护数据一致性
|
||||
|
||||
### 3. 缺少主从复制
|
||||
- 需要配置 PostgreSQL 流复制（Streaming Replication）
|
||||
- 需要配置主从切换机制
|
||||
- 需要使用专门的 PostgreSQL Operator
|
||||
|
||||
## 正确的 PostgreSQL 扩展方案
|
||||
|
||||
### 方案 1: 使用 PostgreSQL Operator
|
||||
|
||||
推荐使用专业的 PostgreSQL Operator：
|
||||
|
||||
#### Zalando PostgreSQL Operator
|
||||
```bash
|
||||
# 添加 Helm 仓库
|
||||
helm repo add postgres-operator-charts https://opensource.zalando.com/postgres-operator/charts/postgres-operator
|
||||
|
||||
# 安装 Operator
|
||||
helm install postgres-operator postgres-operator-charts/postgres-operator
|
||||
|
||||
# 创建 PostgreSQL 集群
|
||||
apiVersion: "acid.zalan.do/v1"
|
||||
kind: postgresql
|
||||
metadata:
|
||||
name: acid-minimal-cluster
|
||||
spec:
|
||||
teamId: "acid"
|
||||
volume:
|
||||
size: 10Gi
|
||||
storageClass: longhorn
|
||||
numberOfInstances: 3 # 1 主 + 2 从
|
||||
users:
|
||||
zalando:
|
||||
- superuser
|
||||
- createdb
|
||||
databases:
|
||||
foo: zalando
|
||||
postgresql:
|
||||
version: "16"
|
||||
```
|
||||
|
||||
#### CloudNativePG Operator
|
||||
```bash
|
||||
# 安装 CloudNativePG
|
||||
kubectl apply -f https://raw.githubusercontent.com/cloudnative-pg/cloudnative-pg/release-1.22/releases/cnpg-1.22.0.yaml
|
||||
|
||||
# 创建集群
|
||||
apiVersion: postgresql.cnpg.io/v1
|
||||
kind: Cluster
|
||||
metadata:
|
||||
name: cluster-example
|
||||
spec:
|
||||
instances: 3
|
||||
storage:
|
||||
storageClass: longhorn
|
||||
size: 10Gi
|
||||
```
|
||||
|
||||
### 方案 2: 读写分离 + KEDA
|
||||
|
||||
如果需要使用 KEDA，正确的架构是：
|
||||
|
||||
```
|
||||
┌─────────────────┐
|
||||
│ 主库 (Master) │ ← 固定 1 个副本，处理写入
|
||||
│ StatefulSet │
|
||||
└─────────────────┘
|
||||
│
|
||||
│ 流复制
|
||||
↓
|
||||
┌─────────────────┐
|
||||
│ 从库 (Replica) │ ← KEDA 管理，处理只读查询
|
||||
│ Deployment │ 可以 0-N 个副本
|
||||
└─────────────────┘
|
||||
```
|
||||
|
||||
**配置示例：**
|
||||
|
||||
```yaml
|
||||
# 主库 - 固定副本
|
||||
apiVersion: apps/v1
|
||||
kind: StatefulSet
|
||||
metadata:
|
||||
name: postgresql-master
|
||||
spec:
|
||||
replicas: 1 # 固定 1 个
|
||||
# ... 配置主库
|
||||
|
||||
---
|
||||
# 从库 - KEDA 管理
|
||||
apiVersion: apps/v1
|
||||
kind: Deployment
|
||||
metadata:
|
||||
name: postgresql-replica
|
||||
spec:
|
||||
# replicas 由 KEDA 管理
|
||||
# ... 配置从库（只读）
|
||||
|
||||
---
|
||||
# KEDA ScaledObject - 只扩展从库
|
||||
apiVersion: keda.sh/v1alpha1
|
||||
kind: ScaledObject
|
||||
metadata:
|
||||
name: postgresql-replica-scaler
|
||||
spec:
|
||||
scaleTargetRef:
|
||||
name: postgresql-replica # 只针对从库
|
||||
minReplicaCount: 0
|
||||
maxReplicaCount: 5
|
||||
triggers:
|
||||
- type: postgresql
|
||||
metadata:
|
||||
connectionString: postgresql://user:pass@postgresql-master:5432/db
|
||||
query: "SELECT COUNT(*) FROM pg_stat_activity WHERE state = 'active' AND query NOT LIKE '%pg_stat_activity%'"
|
||||
targetQueryValue: "10"
|
||||
```
|
||||
|
||||
### 方案 3: 垂直扩展（推荐用于单实例）
|
||||
|
||||
对于单实例 PostgreSQL，使用 VPA (Vertical Pod Autoscaler) 更合适：
|
||||
|
||||
```yaml
|
||||
apiVersion: autoscaling.k8s.io/v1
|
||||
kind: VerticalPodAutoscaler
|
||||
metadata:
|
||||
name: postgresql-vpa
|
||||
spec:
|
||||
targetRef:
|
||||
apiVersion: "apps/v1"
|
||||
kind: StatefulSet
|
||||
name: postgresql
|
||||
updatePolicy:
|
||||
updateMode: "Auto"
|
||||
resourcePolicy:
|
||||
containerPolicies:
|
||||
- containerName: postgresql
|
||||
minAllowed:
|
||||
cpu: 250m
|
||||
memory: 512Mi
|
||||
maxAllowed:
|
||||
cpu: 2000m
|
||||
memory: 4Gi
|
||||
```
|
||||
|
||||
## 当前部署建议
|
||||
|
||||
对于您当前的 PostgreSQL 部署（`/home/fei/k3s/010-中间件/002-postgresql/`）：
|
||||
|
||||
### ❌ 不要使用 KEDA 水平扩展
|
||||
- 当前是单实例 StatefulSet
|
||||
- 没有配置主从复制
|
||||
- 直接扩展会导致数据问题
|
||||
|
||||
### ✅ 推荐的优化方案
|
||||
|
||||
1. **保持单实例运行**
|
||||
```yaml
|
||||
replicas: 1 # 固定不变
|
||||
```
|
||||
|
||||
2. **优化资源配置**
|
||||
```yaml
|
||||
resources:
|
||||
requests:
|
||||
cpu: 500m
|
||||
memory: 1Gi
|
||||
limits:
|
||||
cpu: 2000m
|
||||
memory: 4Gi
|
||||
```
|
||||
|
||||
3. **配置连接池**
|
||||
- 使用 PgBouncer 作为连接池
|
||||
- PgBouncer 可以使用 KEDA 扩展
|
||||
|
||||
4. **定期备份**
|
||||
- 使用 Longhorn 快照
|
||||
- 备份到 S3
|
||||
|
||||
## PgBouncer + KEDA 方案
|
||||
|
||||
这是最实用的方案：PostgreSQL 保持单实例，PgBouncer 使用 KEDA 扩展。
|
||||
|
||||
```yaml
|
||||
# PostgreSQL - 固定单实例
|
||||
apiVersion: apps/v1
|
||||
kind: StatefulSet
|
||||
metadata:
|
||||
name: postgresql
|
||||
spec:
|
||||
replicas: 1 # 固定
|
||||
# ...
|
||||
|
||||
---
|
||||
# PgBouncer - 连接池
|
||||
apiVersion: apps/v1
|
||||
kind: Deployment
|
||||
metadata:
|
||||
name: pgbouncer
|
||||
spec:
|
||||
# replicas 由 KEDA 管理
|
||||
template:
|
||||
spec:
|
||||
containers:
|
||||
- name: pgbouncer
|
||||
image: pgbouncer/pgbouncer:latest
|
||||
# ...
|
||||
|
||||
---
|
||||
# KEDA ScaledObject - 扩展 PgBouncer
|
||||
apiVersion: keda.sh/v1alpha1
|
||||
kind: ScaledObject
|
||||
metadata:
|
||||
name: pgbouncer-scaler
|
||||
spec:
|
||||
scaleTargetRef:
|
||||
name: pgbouncer
|
||||
minReplicaCount: 1
|
||||
maxReplicaCount: 10
|
||||
triggers:
|
||||
- type: postgresql
|
||||
metadata:
|
||||
connectionString: postgresql://postgres:postgres123@postgresql:5432/postgres
|
||||
query: "SELECT COUNT(*) FROM pg_stat_activity WHERE state = 'active'"
|
||||
targetQueryValue: "20"
|
||||
```
|
||||
|
||||
## 总结
|
||||
|
||||
| 方案 | 适用场景 | 复杂度 | 推荐度 |
|
||||
|------|---------|--------|--------|
|
||||
| PostgreSQL Operator | 生产环境，需要高可用 | 高 | ⭐⭐⭐⭐⭐ |
|
||||
| 读写分离 + KEDA | 读多写少场景 | 中 | ⭐⭐⭐⭐ |
|
||||
| PgBouncer + KEDA | 连接数波动大 | 低 | ⭐⭐⭐⭐⭐ |
|
||||
| VPA 垂直扩展 | 单实例，资源需求变化 | 低 | ⭐⭐⭐ |
|
||||
| 直接 KEDA 扩展 | ❌ 不适用 | - | ❌ |
|
||||
|
||||
**对于当前部署，建议保持 PostgreSQL 单实例运行，不使用 KEDA 扩展。**
|
||||
|
||||
如果需要扩展能力，优先考虑：
|
||||
1. 部署 PgBouncer 连接池 + KEDA
|
||||
2. 或者迁移到 PostgreSQL Operator
|
||||
|
||||
---
|
||||
|
||||
**重要提醒：有状态服务的扩展需要特殊处理，不能简单地增加副本数！** ⚠️
|
||||
23
009-基础设施/007-keda/scalers/redis-scaler.yaml
Normal file
23
009-基础设施/007-keda/scalers/redis-scaler.yaml
Normal file
@@ -0,0 +1,23 @@
|
||||
apiVersion: keda.sh/v1alpha1
|
||||
kind: ScaledObject
|
||||
metadata:
|
||||
name: redis-scaler
|
||||
namespace: redis
|
||||
spec:
|
||||
scaleTargetRef:
|
||||
name: redis
|
||||
minReplicaCount: 0 # 空闲时缩容到 0
|
||||
maxReplicaCount: 5 # 最多 5 个副本
|
||||
pollingInterval: 30 # 每 30 秒检查一次
|
||||
cooldownPeriod: 300 # 缩容冷却期 5 分钟
|
||||
triggers:
|
||||
- type: prometheus
|
||||
metadata:
|
||||
serverAddress: http://kube-prometheus-stack-prometheus.monitoring.svc.cluster.local:9090
|
||||
metricName: redis_connected_clients
|
||||
query: sum(redis_connected_clients{namespace="redis"})
|
||||
threshold: "1" # 有连接时启动
|
||||
- type: cpu
|
||||
metricType: Utilization
|
||||
metadata:
|
||||
value: "70" # CPU 使用率超过 70% 时扩容
|
||||
Reference in New Issue
Block a user