fix(deploy): nginx 上游运行期重新解析 + 就绪探针；内存上限按 8C/32G 重新分配

问题：容器重建换 IP 后，网关与 admin-ui 仍往旧 IP 打，/openapi 与 /api
整片 502，必须手动重启 nginx 容器才恢复（2026-11-03 实机踩到）。

编排
- admin-svr / openapi / admin-ui / docs 增加 healthcheck
- admin-ui、gateway、nginx 改 depends_on: condition: service_healthy，
  保证 nginx 启动时后端已就绪
- 新增 deploy/compose.memlimit.yml（原来只存在于某台演示机上）：
  全套上限 8.5G、实测常驻 2G，留给同机其它服务 20G+；
  MySQL buffer pool 128M→768M、max_connections 151→300

nginx
- gateway / admin-ui / iotlcm.conf 示例 统一改 resolver 127.0.0.11 valid=10s
  + upstream zone/resolve：运行期跟随 IP 变化，不必重启（实测 6s 内自愈）

deploy/docker-compose.yml

@@ -175,6 +175,14 @@ services:
         condition: service_healthy
       rabbitmq:
         condition: service_healthy
+    # 就绪探针：nginx 类容器只在启动时解析一次后端 IP，
+    # 没有它就无法用 depends_on 的 service_healthy 保证启动顺序（见 gateway）
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--output-document=/dev/null", "http://127.0.0.1:8008/api.json"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 20s
 
   admin-ui:
     image: 'crpi-qflrqrxoohlkkvgc.cn-chengdu.personal.cr.aliyuncs.com/iotlcm/iotlcm:iotlcm-admin-ui-v1.0.4'
@@ -191,7 +199,14 @@ services:
     networks:
       - iotlcm-network
     depends_on:
-      - admin-svr
+      admin-svr:
+        condition: service_healthy
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--output-document=/dev/null", "http://127.0.0.1/"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 5s
 
   openapi:
     image: 'crpi-qflrqrxoohlkkvgc.cn-chengdu.personal.cr.aliyuncs.com/iotlcm/iotlcm:iotlcm-openapi-v1.0.4'
@@ -213,6 +228,12 @@ services:
         condition: service_healthy
       rabbitmq:
         condition: service_healthy
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--output-document=/dev/null", "http://127.0.0.1:8000/api.json"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 20s
 
   upstream-chan:
     image: 'crpi-qflrqrxoohlkkvgc.cn-chengdu.personal.cr.aliyuncs.com/iotlcm/iotlcm:iotlcm-upstream-chan-v1.0.4'
@@ -370,6 +391,12 @@ services:
       - "127.0.0.1:8019:80"
     networks:
       - iotlcm-network
+    healthcheck:
+      test: ["CMD", "wget", "--quiet", "--output-document=/dev/null", "http://127.0.0.1/docs/"]
+      interval: 10s
+      timeout: 3s
+      retries: 5
+      start_period: 10s
 
   # ---------------- 可选：运营商接口挡板 ----------------
   # 模拟中国移动 OneLink 接口，用于在没有真实运营商环境时联调整条链路：
@@ -418,10 +445,14 @@ services:
     networks:
       - iotlcm-network
     depends_on:
-      - admin-ui
-      - openapi
-      - app-web
-      - docs
+      admin-ui:
+        condition: service_healthy
+      openapi:
+        condition: service_healthy
+      docs:
+        condition: service_healthy
+      app-web:
+        condition: service_started
     profiles:
       - with-nginx
 
@@ -437,10 +468,16 @@ services:
       - ./data/logs/gateway:/var/log/nginx
     networks:
       - iotlcm-network
+    # 后端必须"健康"后才起网关：否则 nginx 会在后端拿到 IP 之前把 upstream
+    # 解析结果缓存下来，容器重启换 IP 后所有请求变成 502（2026-11-03 踩过）。
+    # 配置里另有 resolver + resolve，可在运行期自动跟随 IP 变化。
     depends_on:
-      - admin-ui
-      - openapi
-      - docs
+      admin-ui:
+        condition: service_healthy
+      openapi:
+        condition: service_healthy
+      docs:
+        condition: service_healthy
     profiles:
       - with-gateway