服务器降本增效实践

# 本机成本自查：找出长期低负载却开着高规格的迹象
uptime                                   # 看 15 分钟平均负载
free -g && nproc
sar -r 1 5 2>/dev/null || vmstat 1 5

# 统计过去 30 天 CPU / 内存的真实峰值（需已部署 node_exporter + Prometheus）
# CPU 峰值
# max_over_time((100 - avg by(instance)(rate(node_cpu_seconds_total{mode="idle"}[5m])) * 100)[30d:5m])
# 内存峰值
# max_over_time((100 * (1 - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes))[30d:5m])

# Linux 主机侧：找出 30 天未被访问的站点目录（判断是否可下线）
for d in /var/www/*/; do
  last=$(find "$d" -type f -name '*.log' -mtime -30 | head -1)
  [ -z "$last" ] && echo "疑似闲置站点: $d"
done

# 找出长期无连接的监听端口（服务可能已无人使用）
ss -lntp
# 配合 Nginx 访问日志统计各域名的近 30 天请求数
awk -F'"' '{print $2}' /var/log/nginx/access.log | awk '{print $2}' | sort | uniq -c | sort -rn | head -20

# 找出大文件与长期未访问的冷数据（准备降冷存储）
find /data -type f -size +100M -atime +90 -printf '%s\t%p\n' | sort -rn | head -30

# 降配前的性能基线记录（降配后对比，确认没有劣化）
echo "=== $(date) ===" >> /var/log/baseline.log
uptime >> /var/log/baseline.log
free -m >> /var/log/baseline.log
iostat -x 1 3 >> /var/log/baseline.log 2>/dev/null
curl -sS -o /dev/null -w 'P99探测 http=%{http_code} total=%{time_total}s\n' https://www.example.com/

{
  "Rules": [
    {
      "ID": "logs-tiering",
      "Status": "Enabled",
      "Filter": { "Prefix": "logs/" },
      "Transitions": [
        { "Days": 30, "StorageClass": "STANDARD_IA" },
        { "Days": 90, "StorageClass": "ARCHIVE" }
      ],
      "Expiration": { "Days": 365 }
    },
    {
      "ID": "abort-incomplete-multipart",
      "Status": "Enabled",
      "Filter": { "Prefix": "" },
      "AbortIncompleteMultipartUpload": { "DaysAfterInitiation": 7 }
    }
  ]
}

# 用 AWS CLI 应用生命周期规则（其他云用对应 CLI，参数基本一致）
aws s3api put-bucket-lifecycle-configuration \
  --bucket my-logs-bucket \
  --lifecycle-configuration file://lifecycle.json

aws s3api get-bucket-lifecycle-configuration --bucket my-logs-bucket

# /etc/nginx/conf.d/compress.conf
# Brotli（需 ngx_brotli 模块）优先，不支持时回落 gzip
brotli on;
brotli_comp_level 5;
brotli_min_length 1024;
brotli_types text/plain text/css application/javascript application/json
             application/xml image/svg+xml;

gzip on;
gzip_vary on;
gzip_comp_level 5;
gzip_min_length 1024;
gzip_proxied any;
gzip_types text/plain text/css application/javascript application/json
           application/xml+rss image/svg+xml;

# 静态资源长缓存 + 协商缓存策略
map $uri $cache_control {
    default               "no-cache";
    ~*\.(js|css|png|jpg|jpeg|gif|webp|avif|woff2)$  "public, max-age=31536000, immutable";
}
add_header Cache-Control $cache_control;

# 检查压缩是否真的生效
curl -H 'Accept-Encoding: br' -sSI https://www.example.com/app.js | grep -i 'content-encoding\|content-length'
curl -H 'Accept-Encoding: gzip' -sSI https://www.example.com/app.js | grep -i 'content-encoding\|content-length'

# 统计 CDN 命中率（看响应头）
curl -sSI https://www.example.com/static/logo.png | grep -i 'x-cache\|age\|via'

# 方案 A：用 crontab + 云 CLI 做定时升降配（最通用）
# 每天 22:00 降配（业务低峰），次日 08:30 恢复
0 22 * * * /usr/local/bin/aliyun_cloudops.sh down >> /var/log/autoscale.log 2>&1
30 8  * * * /usr/local/bin/aliyun_cloudops.sh up   >> /var/log/autoscale.log 2>&1

# /usr/local/bin/aliyun_cloudops.sh
#!/usr/bin/env bash
set -euo pipefail
INSTANCES=("i-xxxxxxxxxxxx" "i-yyyyyyyyyyyy")
case "$1" in
  up)   TYPE="ecs.c7.large";   ACTION="升配" ;;
  down) TYPE="ecs.c7.medium";  ACTION="降配" ;;
  *) echo "usage: $0 up|down"; exit 1 ;;
esac
for id in "${INSTANCES[@]}"; do
  echo "$(date '+%F %T') $ACTION $id -> $TYPE"
  # 云平台 CLI 示例（以实际厂商命令为准，先停机再改规格）
  # aliyun ecs ModifyInstanceSpec --InstanceId "$id" --InstanceType "$TYPE"
done

# 方案 B：Kubernetes HPA（容器化场景）
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: web-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: web
  minReplicas: 2
  maxReplicas: 10
  behavior:
    scaleDown:
      stabilizationWindowSeconds: 300   # 缩容冷却 5 分钟，避免抖动
      policies:
        - type: Percent
          value: 20
          periodSeconds: 60
  metrics:
    - type: Resource
      resource:
        name: cpu
        target: { type: Utilization, averageUtilization: 60 }
    - type: Resource
      resource:
        name: memory
        target: { type: Utilization, averageUtilization: 70 }