Docker + Nginx 实现 Java 服务零停机发布并解决发布502问题

nginx 502 Bad Gateway

upstream ring_servers {
    server ring:8080;
    server ring_temp:8080 backup;
}

curl 检查接口返回 000，脚本不动了

upstream ring_servers {
    least_conn;

    # 主服务
    server ring:8080 max_fails=1 fail_timeout=1s;
    # 临时服务（发布兜底，backup 模式）
    server ring_temp:8080 max_fails=3 fail_timeout=30s backup;

    # 核心：出错自动重试下一个节点，用户看不到 502
    proxy_next_upstream error timeout invalid_header http_500 http_502 http_503 http_504;
    proxy_next_upstream_tries 2;
    keepalive 32;
}

proxy_connect_timeout 10s;
proxy_send_timeout 10s;
proxy_read_timeout 30s;
proxy_buffering on;

version: "3.9"
services:
  # 主服务
  ring:
    image: ring:1.0.0
    restart: always
    ports:
      - "8080:8080"
    networks:
      - webnet
    healthcheck:
      test: ["CMD", "curl", "-s", "-o", "/dev/null", "-w", "%{http_code}", "http://localhost:8080/api/health"]
      interval: 5s
      timeout: 3s
      retries: 10
      start_period: 60s
  # 临时服务（发布用）
  ring_temp:
    image: ring:1.0.0
    restart: "no"  # 不自动重启
    ports:
      - "8081:8080"
    networks:
      - webnet
networks:
  webnet:
    driver: bridge

/api/app/checkServerStatus

#!/bin/bash
set +e
# ==================== 配置项（根据自己项目修改） ====================
# 公网要监控的接口（验证用户是否能正常访问）
MONITOR_URL="https://你的域名/api/app/checkServerStatus"
# 主服务端口检查（不经过 Nginx）
MAIN_CHECK_URL="http://127.0.0.1:8080/api/app/checkServerStatus"
# 临时服务端口检查
TEMP_CHECK_URL="http://127.0.0.1:8081/api/app/checkServerStatus"
MAX_RETRY=30
RETRY_INTERVAL=5
# ====================================================================
# 后台实时监控公网接口
start_monitor() {
    while true; do
        current_time=$(date "+%Y-%m-%d %H:%M:%S")
        HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" --insecure $MONITOR_URL 2>/dev/null)
        if [ "$HTTP_CODE" = "200" ]; then
            echo -e "\033[32m[${current_time}] ✅ 公网接口正常：200\033[0m"
        else
            echo -e "\033[31m[${current_time}] ❌ 公网接口异常：${HTTP_CODE}\033[0m"
        fi
        sleep 1
    done
}
# 检查端口服务是否就绪
check_service_ready() {
    local url=$1
    local desc=$2
    echo -e "\033[36m=== 等待 ${desc} 就绪 ===\033[0m"
    local retry=0
    while [ $retry -lt $MAX_RETRY ]; do
        code=$(curl -s -o /dev/null -w "%{http_code}" --insecure $url 2>/dev/null)
        if [ "$code" = "200" ]; then
            echo -e "\033[32m✅ ${desc} 启动成功\033[0m"
            return 0
        fi
        retry=$((retry+1))
        echo "重试 ${retry}/${MAX_RETRY}，状态码：$[code]，等待 ${RETRY_INTERVAL}s..."
        sleep $RETRY_INTERVAL
    done
    echo -e "\033[31m❌ ${desc} 启动超时\033[0m"
    kill $MONITOR_PID 2>/dev/null
    exit 1
}
# ==================== 发布主流程 ====================
echo -e "\n==================== 开始零停机发布 ====================\n"
# 1. 启动后台监控
start_monitor &
MONITOR_PID=$!
# 2. 启动临时副本
echo -e "\n1. 启动临时服务"
docker compose up -d ring_temp
check_service_ready "$TEMP_CHECK_URL" "临时服务"
# 3. 重启主服务
echo -e "\n2. 重启主服务"
docker compose up -d --force-recreate ring
check_service_ready "$MAIN_CHECK_URL" "主服务"
# 4. 关闭临时服务
echo -e "\n3. 关闭临时服务"
docker compose stop ring_temp
docker compose rm -f ring_temp
# 5. 停止监控
kill $MONITOR_PID 2>/dev/null
sleep 1
echo -e "\n\033[32m=====================================================\033[0m"
echo -e "\033[32m✅ 发布完成！全程无 502，用户无感知\033[0m"
echo -e "\033[32m=====================================================\033[0m"
exit 0

chmod +x deploy_ring.sh

./deploy_ring.sh