完成a6000测试系统构建、部署、测试整合 (#35)

测试方案： - lm2机器端口映射到本机：18080, 18081, 8082-8085 - 访问URL: http://localhost:18080/dashboard ![image.png](/attachments/30ed6e20-697a-4d3b-a6d3-6acccd2e9922) ![image.png](/attachments/38ef1751-0f3b-49c6-9100-f70d15617acc) ![image.png](/attachments/3be45005-9b9e-4165-8ef6-1d27405800f1) ![image.png](/attachments/eb916192-edc1-4096-8f9f-9769ab6d9039) ![image.png](/attachments/620e6efc-bd02-45ae-bba1-99a95a1b4c02) ![image.png](/attachments/986e77e7-c687-405f-a760-93282249f72f) 端到端测试通过： ![image.png](/attachments/c6e29875-4a16-4718-8b2f-368f64eb545e) Co-authored-by: sundapeng.sdp <sundapeng@hashdata.cn> Reviewed-on: #35 Reviewed-by: xuxt <xuxt@zgclab.edu.cn> Reviewed-by: sundapeng <sundp@mail.zgclab.edu.cn> Reviewed-by: huhy <husteryezi@163.com>
2025-10-29 10:04:27 +08:00
parent d1b89c0cf6
commit 59a38513a4
143 changed files with 10840 additions and 609 deletions
--- a/src/sys/tests/README.md
+++ b/src/sys/tests/README.md
@@ -32,7 +32,7 @@

 - 一键执行
  - `cd src/sys/tests`
-  - `./scripts/00_e2e_test.sh`
+  - `./scripts/00_e2e_test.sh`（CPU-only）或 `./scripts/00_e2e_test.sh --enable-gpu`（启用 GPU 流程）

 - 分步执行（推荐用于排查）
  - `./scripts/01_bootstrap.sh` 生成目录/拷贝 `update-dns.sh`/构建 agent 二进制/写 `.env`
@@ -42,7 +42,12 @@
  - `./scripts/05_agent_register.sh` 获取两个节点的 `node_id` 与初始 IP，检查本地 `node.json`
  - `./scripts/06_write_health_and_assert.sh` 写健康文件并断言 `nodes.json` 仅包含 2 个在线节点
  - `./scripts/07_logs_send_and_assert.sh` 向两个节点写日志，断言 ES `train-*`/`infer-*` 计数增长
-  - `./scripts/08_restart_agent_reregister.sh` `node-b` 改为固定 IP `172.29.0.200`，验证保持同一节点 ID 且 IP/时间戳更新
+  - `./scripts/08_restart_agent_reregister.sh` `node-b` 改为固定 IP `172.31.0.200`，验证保持同一节点 ID 且 IP/时间戳更新
+  - `./scripts/10_metric_publish.sh` 发布 metric 客户端包到 FTP
+  - `./scripts/11_metric_node_install.sh` 在 CPU 节点安装并验证端点
+  - `./scripts/12_metric_gpu_install.sh` 在 GPU 节点安装并等待 9100/9400 就绪（仅启用 GPU 时）
+  - `./scripts/13_metric_verify.sh` 对 master/Prometheus/数据面/Grafana 做综合校验（含 GPU 时校验 dcgm 指标）
+  - `./scripts/14_metric_cleanup.sh` 清理 FTP 产物
  - `./scripts/09_down.sh` 回收容器、网络并清理 `private*/`、`tmp/`

 - 重置环境
@@ -53,16 +58,17 @@
 ## 二、测试部署架构（docker-compose）

 - 网络
-  - 自定义 bridge：`argus-sys-net`，子网 `172.29.0.0/16`
-  - 固定地址：bind=`172.29.0.2`，master=`172.29.0.10`
+  - 自定义 bridge：`argus-sys-net`，子网 `172.31.0.0/16`
+  - 固定地址：bind=`172.31.0.2`，master=`172.31.0.10`

- 服务与端口
+- 服务与端口（宿主机映射端口由 `01_bootstrap.sh` 自动分配并写入 `.env`）
+  - 关键变量：`MASTER_PORT`、`ES_HTTP_PORT`、`KIBANA_PORT`、`NODE_A_PORT`、`NODE_B_PORT`、`PROMETHEUS_PORT`、`GRAFANA_PORT`、`ALERTMANAGER_PORT`、`WEB_PROXY_PORT_8080..8085`、`FTP_PORT`、`FTP_DATA_PORT`、`FTP_PASSIVE_HOST_RANGE`
  - `bind`（`argus-bind9:latest`）：监听 53/tcp+udp；负责同步 `*.argus.com` 记录
-  - `master`（`argus-master:latest`）：对外 `32300→3000`；API `http://localhost:32300`
-  - `es`（`argus-elasticsearch:latest`）：`9200→9200`；单节点，无安全
-  - `kibana`（`argus-kibana:latest`）：`5601→5601`；通过 `ELASTICSEARCH_HOSTS=http://es:9200` 访问 ES
-  - `node-a`（`ubuntu:22.04`）：同时运行 Fluent Bit + argus-agent，`hostname=dev-yyrshare-nbnyx10-cp2f-pod-0`，`2020→2020`
-  - `node-b`（`ubuntu:22.04`）：同时运行 Fluent Bit + argus-agent，`hostname=dev-yyrshare-uuuu10-ep2f-pod-0`，`2021→2020`
+  - `master`（`argus-master:latest`）：对外 `${MASTER_PORT}→3000`；API `http://localhost:${MASTER_PORT}`
+  - `es`（`argus-elasticsearch:latest`）：`${ES_HTTP_PORT}→9200`；单节点，无安全
+  - `kibana`（`argus-kibana:latest`）：`${KIBANA_PORT}→5601`；通过 `ELASTICSEARCH_HOSTS=http://es:9200` 访问 ES
+  - `node-a`（`ubuntu:22.04`）：同时运行 Fluent Bit + argus-agent，`hostname=dev-yyrshare-nbnyx10-cp2f-pod-0`，`${NODE_A_PORT}→2020`
+  - `node-b`（`ubuntu:22.04`）：同时运行 Fluent Bit + argus-agent，`hostname=dev-yyrshare-uuuu10-ep2f-pod-0`，`${NODE_B_PORT}→2020`

 - 卷与目录
  - 核心服务（bind/master/es/kibana）共享宿主 `./private` 挂载到容器 `/private`
@@ -72,7 +78,7 @@
  - 节点容器的 Fluent Bit/agent 资产以只读方式挂载到 `/assets`/`/usr/local/bin/argus-agent`

 - DNS 配置
-  - 节点容器通过 compose 配置 `dns: [172.29.0.2]` 指向 bind，不挂载 `/etc/resolv.conf`，也不依赖 `update-dns.sh`
+  - 节点容器通过 compose 配置 `dns: [172.31.0.2]` 指向 bind，不挂载 `/etc/resolv.conf`，也不依赖 `update-dns.sh`
  - master/es/kibana 仍共享 `./private`，master 启动会写 `/private/argus/etc/master.argus.com` 供 bind 同步 A 记录

 - 节点入口
@@ -106,6 +112,7 @@
  - 判定：
    - `private/argus/etc/master.argus.com` 存在且为 master IP
    - 在 node-a/node-b 内 `getent hosts master.argus.com` 成功解析到 master IP
+    - 在 metric CPU/GPU 节点内可解析 `master.argus.com` 与 `prom.metric.argus.com`

 - `05_agent_register.sh`
  - 目的：确认两个节点注册到 master 并持久化 `node.json`
@@ -136,3 +143,16 @@
 ---

 如需更严格的断言（例如 Kibana 载入具体插件、ES 文档字段校验），可在 `07_*.sh` 中追加查询与校验逻辑。
+
+---
+
+## 可选：GPU 流程说明
+- 前置条件：宿主安装 NVIDIA 驱动与 `nvidia-container-toolkit`，`nvidia-smi` 在宿主可用。
+- 启用方式：
+  - 一键：`./scripts/00_e2e_test.sh --enable-gpu`
+  - 分步：设置 `ARGUS_SYS_ENABLE_GPU=true` 后执行 `01_bootstrap.sh`、`02_up.sh`；或直接在 `.env` 中将 `ENABLE_GPU=true` 后单独运行 `02_up.sh`。
+- `01_bootstrap.sh` 会写入：
+  - `METRIC_TEST_HOSTNAME_GPU=test-metric-gpu-node-001`
+  - `METRIC_TEST_INSTANCE_GPU=172.31.0.51:9100`
+  - `METRIC_TEST_DCGM_GPU=172.31.0.51:9400`
+- 验证点：`04_verify_dns_routing.sh` 增加对 metric 节点的域名解析；`12_metric_gpu_install.sh` 等待 9100/9400；`13_metric_verify_*` 校验 dcgm 指标与 Grafana 面板。
--- a/src/sys/tests/docker-compose.yml
+++ b/src/sys/tests/docker-compose.yml
@@ -1,21 +1,18 @@
-version: "3.8"
-
 networks:
-  default:
-    name: argus-sys-net
+  sysnet:
    driver: bridge
    ipam:
      driver: default
      config:
-        - subnet: 172.29.0.0/16
+        - subnet: 172.31.0.0/16

 services:
  bind:
    image: ${BIND_IMAGE_TAG:-argus-bind9:latest}
    container_name: argus-bind-sys
    networks:
-      default:
-        ipv4_address: 172.29.0.2
+      sysnet:
+        ipv4_address: 172.31.0.2
    volumes:
      - ./private:/private
    restart: unless-stopped
@@ -32,14 +29,14 @@ services:
      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
    ports:
-      - "32300:3000"
+      - "${MASTER_PORT:-32300}:3000"
    volumes:
      - ./private/argus/master:/private/argus/master
      - ./private/argus/metric/prometheus:/private/argus/metric/prometheus
      - ./private/argus/etc:/private/argus/etc
    networks:
-      default:
-        ipv4_address: 172.29.0.10
+      sysnet:
+        ipv4_address: 172.31.0.10
    restart: unless-stopped

  es:
@@ -55,8 +52,11 @@ services:
      - ./private/argus/log/elasticsearch:/private/argus/log/elasticsearch
      - ./private/argus/etc:/private/argus/etc
    ports:
-      - "9200:9200"
+      - "${ES_HTTP_PORT:-9200}:9200"
    restart: unless-stopped
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.3

  kibana:
    image: argus-kibana:latest
@@ -71,11 +71,14 @@ services:
    depends_on:
      - es
    ports:
-      - "5601:5601"
+      - "${KIBANA_PORT:-5601}:5601"
    restart: unless-stopped
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.4

  node-a:
-    image: ubuntu:22.04
+    image: argus-sys-node:latest
    container_name: argus-node-a
    hostname: dev-yyrshare-nbnyx10-cp2f-pod-0
    depends_on:
@@ -101,13 +104,16 @@ services:
    entrypoint:
      - /usr/local/bin/node-entrypoint.sh
    dns:
-      - 172.29.0.2
+      - 172.31.0.2   # internal bind for *.argus.com
+      - 8.8.8.8      # external fallback for apt/external domains
    ports:
-      - "2020:2020"
+      - "${NODE_A_PORT:-2020}:2020"
    restart: unless-stopped
+    networks:
+      - sysnet

  node-b:
-    image: ubuntu:22.04
+    image: argus-sys-node:latest
    container_name: argus-node-b
    hostname: dev-yyrshare-uuuu10-ep2f-pod-0
    depends_on:
@@ -133,7 +139,269 @@ services:
    entrypoint:
      - /usr/local/bin/node-entrypoint.sh
    dns:
-      - 172.29.0.2
+      - 172.31.0.2
+      - 8.8.8.8
    ports:
-      - "2021:2020"
+      - "${NODE_B_PORT:-2021}:2020"
    restart: unless-stopped
+    networks:
+      - sysnet
+
+  ftp:
+    image: argus-metric-ftp:latest
+    container_name: argus-ftp
+    restart: unless-stopped
+    environment:
+      - TZ=Asia/Shanghai
+      - FTP_BASE_PATH=/private/argus/ftp
+      - FTP_PASSWORD=${FTP_PASSWORD:-ZGClab1234!}
+      - DOMAIN=${FTP_DOMAIN:-ftp.metric.argus.com}
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+    ports:
+      - "${FTP_PORT:-21}:21"
+      - "${FTP_DATA_PORT:-20}:20"
+      - "${FTP_PASSIVE_HOST_RANGE:-21100-21110}:21100-21110"
+    volumes:
+      - ./private/argus/metric/ftp:/private/argus/ftp
+      - ./private/argus/etc:/private/argus/etc
+      - /etc/localtime:/etc/localtime:ro
+      - /etc/timezone:/etc/timezone:ro
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.40
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  prometheus:
+    image: argus-metric-prometheus:latest
+    container_name: argus-prometheus
+    restart: unless-stopped
+    environment:
+      - TZ=Asia/Shanghai
+      - PROMETHEUS_BASE_PATH=/private/argus/metric/prometheus
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+    ports:
+      - "${PROMETHEUS_PORT:-9090}:9090"
+    volumes:
+      - ./private/argus/metric/prometheus:/private/argus/metric/prometheus
+      - ./private/argus/etc:/private/argus/etc
+      - /etc/localtime:/etc/localtime:ro
+      - /etc/timezone:/etc/timezone:ro
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.41
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  grafana:
+    image: argus-metric-grafana:latest
+    container_name: argus-grafana
+    restart: unless-stopped
+    environment:
+      - TZ=Asia/Shanghai
+      - GRAFANA_BASE_PATH=/private/argus/metric/grafana
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+      - GF_SERVER_HTTP_PORT=3000
+      - GF_LOG_LEVEL=warn
+      - GF_LOG_MODE=console
+      - GF_PATHS_PROVISIONING=/private/argus/metric/grafana/provisioning
+      - GF_AUTH_ANONYMOUS_ENABLED=true
+      - GF_AUTH_ANONYMOUS_ORG_ROLE=Viewer
+    ports:
+      - "${GRAFANA_PORT:-3000}:3000"
+    volumes:
+      - ./private/argus/metric/grafana:/private/argus/metric/grafana
+      - ./private/argus/etc:/private/argus/etc
+      - /etc/localtime:/etc/localtime:ro
+      - /etc/timezone:/etc/timezone:ro
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.42
+    depends_on:
+      - prometheus
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  # --- Added: Web Frontend (no host port; resolved by DNS as web.argus.com) ---
+  web-frontend:
+    image: argus-web-frontend:latest
+    container_name: argus-web-frontend
+    environment:
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+      # Frontend runtime-injected external ports (used to render hyperlinks)
+      - EXTERNAL_MASTER_PORT=${WEB_PROXY_PORT_8085:-8085}
+      - EXTERNAL_ALERTMANAGER_PORT=${WEB_PROXY_PORT_8084:-8084}
+      - EXTERNAL_GRAFANA_PORT=${WEB_PROXY_PORT_8081:-8081}
+      - EXTERNAL_PROMETHEUS_PORT=${WEB_PROXY_PORT_8082:-8082}
+      - EXTERNAL_KIBANA_PORT=${WEB_PROXY_PORT_8083:-8083}
+    volumes:
+      - ./private/argus/etc:/private/argus/etc
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.80
+    restart: unless-stopped
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  test-node:
+    image: argus-sys-metric-test-node:latest
+    container_name: argus-metric-test-node
+    hostname: test-metric-node-001
+    restart: unless-stopped
+    privileged: true
+    depends_on:
+      - ftp
+      - prometheus
+    environment:
+      - TZ=Asia/Shanghai
+      - DEBIAN_FRONTEND=noninteractive
+      - FTP_DOMAIN=${FTP_DOMAIN:-ftp.metric.argus.com}
+      - FTP_SERVER=${FTP_SERVER:-172.31.0.40}
+      - FTP_USER=${FTP_USER:-ftpuser}
+      - FTP_PASSWORD=${FTP_PASSWORD:-ZGClab1234!}
+      - FTP_PORT=${FTP_PORT:-21}
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+      - METRIC_NODE_ROLE=cpu
+    volumes:
+      - ./private/argus/agent:/private/argus/agent
+      - ./scripts/metric/test-node-entrypoint.sh:/usr/local/bin/metric-test-node-entrypoint.sh:ro
+      - /etc/localtime:/etc/localtime:ro
+      - /etc/timezone:/etc/timezone:ro
+    entrypoint:
+      - /usr/local/bin/metric-test-node-entrypoint.sh
+    command:
+      - sleep
+      - infinity
+    dns:
+      - 172.31.0.2
+      - 8.8.8.8
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.50
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  test-gpu-node:
+    profiles: ["gpu"]
+    image: argus-sys-metric-test-gpu-node:latest
+    container_name: argus-metric-test-gpu-node
+    hostname: test-metric-gpu-node-001
+    restart: unless-stopped
+    privileged: true
+    runtime: nvidia
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: 
+                - gpu
+    depends_on:
+      - ftp
+      - prometheus
+    environment:
+      - TZ=Asia/Shanghai
+      - DEBIAN_FRONTEND=noninteractive
+      - NVIDIA_VISIBLE_DEVICES=all
+      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
+      - GPU_MODE=gpu
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+      - METRIC_NODE_ROLE=gpu
+    volumes:
+      - ./private/argus/agent:/private/argus/agent
+      - ./scripts/metric/test-node-entrypoint.sh:/usr/local/bin/metric-test-node-entrypoint.sh:ro
+      - /etc/localtime:/etc/localtime:ro
+      - /etc/timezone:/etc/timezone:ro
+    entrypoint:
+      - /usr/local/bin/metric-test-node-entrypoint.sh
+    command:
+      - sleep
+      - infinity
+    dns:
+      - 172.31.0.2
+      - 8.8.8.8
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.51
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  # --- Added: Alertmanager ---
+  alertmanager:
+    image: argus-alertmanager:latest
+    container_name: argus-alertmanager
+    environment:
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+    volumes:
+      - ./private/argus/etc:/private/argus/etc
+      - ./private/argus/alert/alertmanager:/private/argus/alert/alertmanager
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.82
+    ports:
+      - "${ALERTMANAGER_PORT:-9093}:9093"
+    restart: unless-stopped
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
+
+  # --- Added: Web Proxy (multi-port gateway) ---
+  web-proxy:
+    image: argus-web-proxy:latest
+    container_name: argus-web-proxy
+    depends_on:
+      - bind
+      - master
+      - grafana
+      - prometheus
+      - kibana
+      - alertmanager
+    environment:
+      - ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+      - ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+    volumes:
+      - ./private/argus/etc:/private/argus/etc
+    networks:
+      sysnet:
+        ipv4_address: 172.31.0.81
+    ports:
+      - "${WEB_PROXY_PORT_8080:-8080}:8080"
+      - "${WEB_PROXY_PORT_8081:-8081}:8081"
+      - "${WEB_PROXY_PORT_8082:-8082}:8082"
+      - "${WEB_PROXY_PORT_8083:-8083}:8083"
+      - "${WEB_PROXY_PORT_8084:-8084}:8084"
+      - "${WEB_PROXY_PORT_8085:-8085}:8085"
+    restart: unless-stopped
+    logging:
+      driver: "json-file"
+      options:
+        max-size: "10m"
+        max-file: "3"
--- a/src/sys/tests/scripts/00_e2e_test.sh
+++ b/src/sys/tests/scripts/00_e2e_test.sh
@@ -3,6 +3,45 @@ set -euo pipefail

 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"

+ENABLE_GPU=false
+CLEANUP=true
+
+usage() {
+  cat <<'EOF'
+Usage: 00_e2e_test.sh [options]
+
+Options:
+  --enable-gpu   启用 GPU 相关拓扑与测试流程
+  --no-clean     跳过清理流程（不执行 14 和 09）
+  -h, --help     显示帮助信息
+EOF
+}
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --enable-gpu)
+      ENABLE_GPU=true
+      shift
+      ;;
+    --no-clean)
+      CLEANUP=false
+      shift
+      ;;
+    -h|--help)
+      usage
+      exit 0
+      ;;
+    *)
+      echo "Unknown argument: $1" >&2
+      usage
+      exit 1
+      ;;
+  esac
+done
+
+export ARGUS_SYS_ENABLE_GPU=$ENABLE_GPU
+
+# 基础步骤（不包含清理与下线）
 SCRIPTS=(
  "01_bootstrap.sh"
  "02_up.sh"
@@ -12,9 +51,20 @@ SCRIPTS=(
  "06_write_health_and_assert.sh"
  "07_logs_send_and_assert.sh"
  "08_restart_agent_reregister.sh"
-  "09_down.sh"
+  "10_metric_publish.sh"
+  "11_metric_node_install.sh"
+  "12_metric_gpu_install.sh"
+  "13_metric_verify.sh"
 )

+# 如未禁用清理，则追加清理与下线步骤（保持原有顺序）
+if [[ "$CLEANUP" == "true" ]]; then
+  SCRIPTS+=(
+    "14_metric_cleanup.sh"
+    "09_down.sh"
+  )
+fi
+
 for script in "${SCRIPTS[@]}"; do
  echo "[SYS-E2E] Running $script"
  "$SCRIPT_DIR/$script"
@@ -22,5 +72,8 @@ for script in "${SCRIPTS[@]}"; do
  echo
 done

-echo "[SYS-E2E] All tests completed"
-
+if [[ "$CLEANUP" == "true" ]]; then
+  echo "[SYS-E2E] All tests completed"
+else
+  echo "[SYS-E2E] All tests completed (cleanup skipped)"
+fi
--- a/src/sys/tests/scripts/01_bootstrap.sh
+++ b/src/sys/tests/scripts/01_bootstrap.sh
@@ -22,24 +22,66 @@ ensure_image() {
 }

 echo "[INFO] Preparing directories..."
+ensure_writable_dir() {
+  local path="$1"
+  local parent
+  parent="$(dirname "$path")"
+  mkdir -p "$parent" 2>/dev/null || true
+  mkdir -p "$path" 2>/dev/null || true
+  if [[ ! -w "$path" ]]; then
+    docker run --rm -v "$parent:/target" ubuntu:24.04 bash -lc "chown -R $(id -u):$(id -g) /target" >/dev/null 2>&1 || true
+  fi
+  mkdir -p "$path"
+}
+
+# preflight: make base dirs writable if inherited from root-owned mounts
+ensure_writable_dir "$PRIVATE_CORE/argus"
+ensure_writable_dir "$PRIVATE_CORE/argus/metric"
+ensure_writable_dir "$PRIVATE_CORE/argus/metric/grafana"
+ensure_writable_dir "$PRIVATE_CORE/argus/metric/prometheus"
+
 mkdir -p \
  "$PRIVATE_CORE/argus/etc" \
  "$PRIVATE_CORE/argus/bind" \
  "$PRIVATE_CORE/argus/master" \
  "$PRIVATE_CORE/argus/metric/prometheus" \
+  "$PRIVATE_CORE/argus/alert/alertmanager" \
+  "$PRIVATE_CORE/argus/metric/ftp/share" \
+  "$PRIVATE_CORE/argus/metric/grafana/data" \
+  "$PRIVATE_CORE/argus/metric/grafana/logs" \
+  "$PRIVATE_CORE/argus/metric/grafana/plugins" \
+  "$PRIVATE_CORE/argus/metric/grafana/provisioning/datasources" \
+  "$PRIVATE_CORE/argus/metric/grafana/provisioning/dashboards" \
+  "$PRIVATE_CORE/argus/metric/grafana/data/sessions" \
+  "$PRIVATE_CORE/argus/metric/grafana/data/dashboards" \
+  "$PRIVATE_CORE/argus/metric/grafana/config" \
+  "$PRIVATE_CORE/argus/metric/prometheus/data" \
+  "$PRIVATE_CORE/argus/metric/prometheus/rules" \
+  "$PRIVATE_CORE/argus/metric/prometheus/targets" \
+  "$PRIVATE_CORE/argus/agent" \
  "$PRIVATE_CORE/argus/log/elasticsearch" \
  "$PRIVATE_CORE/argus/log/kibana" \
  "$PRIVATE_NODEA/argus/agent/dev-yyrshare-nbnyx10-cp2f-pod-0/health" \
  "$PRIVATE_NODEB/argus/agent/dev-yyrshare-uuuu10-ep2f-pod-0/health" \
  "$TMP_DIR"

-# Align ownership for supervisor-managed services (ES/Kibana expect UID/GID inside container)
+# Align ownership for supervisor-managed services (ES/Kibana/Grafana expect UID/GID inside container)
 echo "[INFO] Fixing ownership for core private directories..."
 chown -R "${ARGUS_BUILD_UID}:${ARGUS_BUILD_GID}" \
  "$PRIVATE_CORE/argus/log/elasticsearch" \
  "$PRIVATE_CORE/argus/log/kibana" \
+  "$PRIVATE_CORE/argus/metric/grafana" \
+  "$PRIVATE_CORE/argus/metric/prometheus" \
+  "$PRIVATE_CORE/argus/alert" \
+  "$PRIVATE_CORE/argus/metric/ftp" \
+  "$PRIVATE_CORE/argus/agent" \
  "$PRIVATE_CORE/argus/etc" 2>/dev/null || true

+# 确保 alert 与 etc 目录组可写，便于非 root 且仅匹配 GID 的服务写入运行文件
+chmod -R g+w "$PRIVATE_CORE/argus/alert" "$PRIVATE_CORE/argus/etc" 2>/dev/null || true
+
+echo "[INFO] Using compose-managed network (auto-created by docker compose)"
+
 echo "[INFO] Distributing update-dns.sh for core services (bind/master/es/kibana)"
 BIND_UPDATE_SRC="$REPO_ROOT/src/bind/build/update-dns.sh"
 BIND_UPDATE_DEST="$PRIVATE_CORE/argus/etc/update-dns.sh"
@@ -55,6 +97,51 @@ ensure_image "argus-elasticsearch:latest"
 ensure_image "argus-kibana:latest"
 ensure_image "argus-bind9:latest"
 ensure_image "argus-master:latest"
+ensure_image "argus-metric-ftp:latest"
+ensure_image "argus-metric-prometheus:latest"
+ensure_image "argus-metric-grafana:latest"
+ensure_image "argus-web-frontend:latest"
+ensure_image "argus-web-proxy:latest"
+ensure_image "argus-alertmanager:latest"
+
+echo "[INFO] Preparing Fluent Bit local dependency packages..."
+FLB_BUILD_PACKAGES_DIR="$REPO_ROOT/src/log/fluent-bit/build/packages"
+mkdir -p "$FLB_BUILD_PACKAGES_DIR"
+for deb in \
+  "$REPO_ROOT/src/metric/client-plugins/all-in-one-full/plugins/fluent-bit/bin/libyaml-0-2_"*_amd64.deb \
+  "$REPO_ROOT/src/metric/client-plugins/all-in-one-full/plugins/fluent-bit/bin/libpq5_"*_amd64.deb ; do
+  if ls $deb >/dev/null 2>&1; then
+    for f in $deb; do
+      base="$(basename "$f")"
+      if [[ ! -f "$FLB_BUILD_PACKAGES_DIR/$base" ]]; then
+        cp "$f" "$FLB_BUILD_PACKAGES_DIR/"
+        echo "  [+] copied $base"
+      fi
+    done
+  fi
+done
+
+# 额外：从 all-in-one-full 的 ubuntu22/curl.tar.gz 解包必要依赖（libsasl2/ldap），便于离线安装
+CURLOPT_TAR="$REPO_ROOT/src/metric/client-plugins/all-in-one-full/deps/ubuntu22/curl.tar.gz"
+if [[ -f "$CURLOPT_TAR" ]]; then
+  tmpdir=$(mktemp -d)
+  if tar -xzf "$CURLOPT_TAR" -C "$tmpdir" 2>/dev/null; then
+    for p in \
+      libsasl2-2_*_amd64.deb \
+      libsasl2-modules-db_*_amd64.deb \
+      libldap-2.5-0_*_amd64.deb \
+      libidn2-0_*_amd64.deb \
+      libbrotli1_*_amd64.deb \
+      libssl3_*_amd64.deb ; do
+      src=$(ls "$tmpdir"/curl/$p 2>/dev/null | head -n1 || true)
+      if [[ -n "$src" ]]; then
+        base="$(basename "$src")"
+        [[ -f "$FLB_BUILD_PACKAGES_DIR/$base" ]] || cp "$src" "$FLB_BUILD_PACKAGES_DIR/" && echo "  [+] staged $base"
+      fi
+    done
+  fi
+  rm -rf "$tmpdir"
+fi

 echo "[INFO] Building agent binary..."
 pushd "$REPO_ROOT/src/agent" >/dev/null
@@ -68,10 +155,139 @@ if [[ ! -x "$AGENT_BIN" ]]; then
 fi
 echo "$AGENT_BIN" > "$TMP_DIR/agent_binary_path"

-echo "[INFO] Writing .env with UID/GID"
+# 检测GPU环境
+REQUEST_GPU=${ARGUS_SYS_ENABLE_GPU:-false}
+GPU_CHECK_SCRIPT="$REPO_ROOT/src/metric/tests/scripts/common/check-gpu.sh"
+if [[ "$REQUEST_GPU" == "true" ]]; then
+  echo "[INFO] --enable-gpu 已启用，开始检测GPU环境..."
+  if [[ -f "$GPU_CHECK_SCRIPT" ]]; then
+    if bash "$GPU_CHECK_SCRIPT" >/dev/null 2>&1; then
+      echo "[INFO] GPU环境可用，将在 compose 中启用 test-gpu-node"
+      GPU_AVAILABLE=true
+    else
+      echo "[ERROR] 未检测到可用 GPU，但指定了 --enable-gpu" >&2
+      exit 1
+    fi
+  else
+    echo "[ERROR] 未找到 GPU 检测脚本: $GPU_CHECK_SCRIPT" >&2
+    exit 1
+  fi
+else
+  GPU_AVAILABLE=false
+  echo "[INFO] GPU 支持未启用，跳过 GPU 检测"
+fi
+
+echo "[INFO] Writing .env with UID/GID and metric configuration"
+#############################################
+# 动态分配宿主机端口并写入 .env
+#############################################
+
+# 读取现有 .env（若存在），用于保留密码/域名等
+EXIST_DOTENV="$TEST_ROOT/.env"
+if [[ -f "$EXIST_DOTENV" ]]; then
+  EXISTING_FTP_PASSWORD="$(grep -E '^FTP_PASSWORD=' "$EXIST_DOTENV" | tail -n1 | sed 's/^FTP_PASSWORD=//')"
+  EXISTING_FTP_DOMAIN="$(grep -E '^FTP_DOMAIN=' "$EXIST_DOTENV" | tail -n1 | sed 's/^FTP_DOMAIN=//')"
+  EXISTING_USE_INTRANET="$(grep -E '^USE_INTRANET=' "$EXIST_DOTENV" | tail -n1 | sed 's/^USE_INTRANET=//')"
+else
+  EXISTING_FTP_PASSWORD=""
+  EXISTING_FTP_DOMAIN=""
+  EXISTING_USE_INTRANET=""
+fi
+
+is_port_free() {
+  local p="$1"
+  ss -ltnH 2>/dev/null | awk -v pat=":${p}$" '$4 ~ pat{f=1} END{exit f?1:0}'
+}
+
+find_free_port() {
+  local prefer="$1"; local start_scan="${2:-20000}"; local max="${3:-65000}"
+  if is_port_free "$prefer"; then echo "$prefer"; return 0; fi
+  local p
+  for (( p=start_scan; p<=max; p++ )); do
+    if is_port_free "$p"; then echo "$p"; return 0; fi
+  done
+  return 1
+}
+
+find_free_range() {
+  local begin="$1"; local end="$2"; local need_count=$((end-begin+1))
+  local try_start="$begin"
+  while (( try_start + need_count - 1 <= 65000 )); do
+    local ok=1
+    for (( p=try_start; p<try_start+need_count; p++ )); do
+      if ! is_port_free "$p"; then ok=0; break; fi
+    done
+    if (( ok==1 )); then
+      echo "${try_start}-$((try_start+need_count-1))"; return 0
+    fi
+    ((try_start++))
+  done
+  return 1
+}
+
+# 优先端口（若占用则回退到 20000+ 的空闲）
+MASTER_PORT_VAL=$(find_free_port 32300)
+ES_HTTP_PORT_VAL=$(find_free_port 9200)
+KIBANA_PORT_VAL=$(find_free_port 5601)
+NODE_A_PORT_VAL=$(find_free_port 2020)
+NODE_B_PORT_VAL=$(find_free_port 2021)
+PROMETHEUS_PORT_VAL=$(find_free_port 9090)
+GRAFANA_PORT_VAL=$(find_free_port 3000)
+ALERTMANAGER_PORT_VAL=$(find_free_port 9093)
+WEB_PROXY_PORT_8080_VAL=$(find_free_port 8080 18080 65000)
+WEB_PROXY_PORT_8081_VAL=$(find_free_port 8081 18081 65000)
+WEB_PROXY_PORT_8082_VAL=$(find_free_port 8082 18082 65000)
+WEB_PROXY_PORT_8083_VAL=$(find_free_port 8083 18083 65000)
+WEB_PROXY_PORT_8084_VAL=$(find_free_port 8084 18084 65000)
+WEB_PROXY_PORT_8085_VAL=$(find_free_port 8085 18085 65000)
+
+# FTP 控制/数据端口：若 21/20 被占用则回退到空闲
+FTP_PORT_VAL=$(find_free_port 21 2121 65000)
+FTP_DATA_PORT_VAL=$(find_free_port 20 2120 65000)
+
+# FTP 被动端口范围：优先 21100-21110，若被占用则向后平移
+FTP_PASV_RANGE_VAL="$(find_free_range 21100 21110 || echo 22100-22110)"
+
+# 变量持久化到 .env
 cat > "$TEST_ROOT/.env" <<EOF
 ARGUS_BUILD_UID=$ARGUS_BUILD_UID
 ARGUS_BUILD_GID=$ARGUS_BUILD_GID
+
+# GPU 配置
+ENABLE_GPU=$GPU_AVAILABLE
+
+# 测试节点（CPU/GPU）默认标识与实例
+METRIC_TEST_HOSTNAME_CPU=test-metric-node-001
+METRIC_TEST_INSTANCE_CPU=172.31.0.50:9100
+METRIC_TEST_HOSTNAME_GPU=test-metric-gpu-node-001
+METRIC_TEST_INSTANCE_GPU=172.31.0.51:9100
+METRIC_TEST_DCGM_GPU=172.31.0.51:9400
+
+# Master/日志/监控等服务的宿主机端口（自动分配）
+MASTER_PORT=$MASTER_PORT_VAL
+ES_HTTP_PORT=$ES_HTTP_PORT_VAL
+KIBANA_PORT=$KIBANA_PORT_VAL
+NODE_A_PORT=$NODE_A_PORT_VAL
+NODE_B_PORT=$NODE_B_PORT_VAL
+PROMETHEUS_PORT=$PROMETHEUS_PORT_VAL
+GRAFANA_PORT=$GRAFANA_PORT_VAL
+ALERTMANAGER_PORT=$ALERTMANAGER_PORT_VAL
+WEB_PROXY_PORT_8080=$WEB_PROXY_PORT_8080_VAL
+WEB_PROXY_PORT_8081=$WEB_PROXY_PORT_8081_VAL
+WEB_PROXY_PORT_8082=$WEB_PROXY_PORT_8082_VAL
+WEB_PROXY_PORT_8083=$WEB_PROXY_PORT_8083_VAL
+WEB_PROXY_PORT_8084=$WEB_PROXY_PORT_8084_VAL
+WEB_PROXY_PORT_8085=$WEB_PROXY_PORT_8085_VAL
+
+# FTP 配置（自动分配：若21/20占用则使用高位端口）
+FTP_PORT=$FTP_PORT_VAL
+FTP_DATA_PORT=$FTP_DATA_PORT_VAL
+FTP_PASSIVE_HOST_RANGE=$FTP_PASV_RANGE_VAL
+FTP_PASSWORD=${EXISTING_FTP_PASSWORD:-ZGClab1234!}
+FTP_DOMAIN=${EXISTING_FTP_DOMAIN:-ftp.metric.argus.com}
+
+# 网络配置
+USE_INTRANET=${EXISTING_USE_INTRANET:-false}
 EOF

-echo "[OK] Bootstrap completed"
+echo "[OK] Bootstrap completed. Ports written to $TEST_ROOT/.env"
--- a/src/sys/tests/scripts/02_up.sh
+++ b/src/sys/tests/scripts/02_up.sh
@@ -3,6 +3,7 @@ set -euo pipefail

 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+REPO_ROOT="$(cd "$TEST_ROOT/../../.." && pwd)"

 compose() {
  if docker compose version >/dev/null 2>&1; then
@@ -13,10 +14,92 @@ compose() {
 }

 echo "[INFO] Bringing up system stack..."
+
+# 加载 .env 以获取端口（由 01_bootstrap 生成）
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+# GPU 开关优先级：显式环境变量 > .env 中的 ENABLE_GPU > 默认 false
+if [[ "${ARGUS_SYS_ENABLE_GPU:-}" == "true" ]]; then
+  REQUEST_GPU=true
+elif [[ "${ARGUS_SYS_ENABLE_GPU:-}" == "false" ]]; then
+  REQUEST_GPU=false
+else
+  REQUEST_GPU=${ENABLE_GPU:-false}
+fi
+
+GPU_AVAILABLE=false
+GPU_CHECK_SCRIPT="$REPO_ROOT/src/metric/tests/scripts/common/check-gpu.sh"
+
+if [[ "$REQUEST_GPU" == "true" ]]; then
+  echo "[INFO] --enable-gpu 生效，验证主机 GPU..."
+  if [[ -f "$GPU_CHECK_SCRIPT" ]]; then
+    if bash "$GPU_CHECK_SCRIPT" >/dev/null 2>&1; then
+      GPU_AVAILABLE=true
+      echo "[INFO] GPU 检测通过，将启动 gpu profile"
+    else
+      echo "[ERROR] 主机缺少可用 GPU，无法继续 --enable-gpu 流程" >&2
+      exit 1
+    fi
+  else
+    echo "[ERROR] 未找到 GPU 检测脚本: $GPU_CHECK_SCRIPT" >&2
+    exit 1
+  fi
+else
+  echo "[INFO] 未启用 GPU 流程"
+fi
+
 pushd "$TEST_ROOT" >/dev/null
 compose -p argus-sys down --remove-orphans || true
-compose -p argus-sys up -d
+
+# 清理可能由 08 脚本创建的同名容器，避免 compose up 冲突
+for name in argus-node-b; do
+  if docker ps -aqf "name=^${name}$" >/dev/null 2>&1 && [[ -n "$(docker ps -aqf "name=^${name}$")" ]]; then
+    docker rm -f "$name" >/dev/null 2>&1 || true
+  fi
+done
+
+# 预检：检查多端口网关所需宿主端口是否空闲
+check_port_free() {
+  local p="$1"
+  if ss -ltnp 2>/dev/null | grep -q ":${p} "; then
+    echo "[ERR] Host port ${p} is already in use. Please free it before running 02_up.sh" >&2
+    ss -ltnp | awk -v p=":${p} " '$0 ~ p {print "    " $0}' || true
+    return 1
+  fi
+  return 0
+}
+
+for port in \
+  "${WEB_PROXY_PORT_8080:-8080}" \
+  "${WEB_PROXY_PORT_8081:-8081}" \
+  "${WEB_PROXY_PORT_8082:-8082}" \
+  "${WEB_PROXY_PORT_8083:-8083}" \
+  "${WEB_PROXY_PORT_8084:-8084}" \
+  "${WEB_PROXY_PORT_8085:-8085}"; do
+  check_port_free "$port" || { echo "[ERR] Required port busy: $port"; exit 1; }
+done
+
+# 根据GPU可用性决定启动的服务
+if [[ "$GPU_AVAILABLE" == true ]]; then
+  echo "[INFO] 启动所有服务（包含 gpu profile）..."
+  compose -p argus-sys --profile gpu up -d || true
+else
+  echo "[INFO] 启动基础服务（不含 gpu profile）..."
+  compose -p argus-sys up -d || true
+fi
+
+# 若 web-proxy 处于 Created 状态，尝试单独启动一次（处理偶发 Address already in use 后端已释放的场景）
+if docker ps -a --format '{{.Names}}\t{{.Status}}' | grep -q '^argus-web-proxy\s\+Created'; then
+  echo "[WARN] web-proxy in Created state; retry starting it..."
+  docker start argus-web-proxy || true
+fi
+
 popd >/dev/null

-echo "[OK] Services started: master:32300 es:9200 kibana:5601 node-a:2020 node-b:2021"
-
+if [[ "$GPU_AVAILABLE" == true ]]; then
+  echo "[OK] Services started: master:${MASTER_PORT:-32300} es:${ES_HTTP_PORT:-9200} kibana:${KIBANA_PORT:-5601} node-a:${NODE_A_PORT:-2020} node-b:${NODE_B_PORT:-2021} test-gpu-node:172.31.0.51"
+else
+  echo "[OK] Services started: master:${MASTER_PORT:-32300} es:${ES_HTTP_PORT:-9200} kibana:${KIBANA_PORT:-5601} node-a:${NODE_A_PORT:-2020} node-b:${NODE_B_PORT:-2021} (gpu skipped)"
+fi
--- a/src/sys/tests/scripts/03_wait_ready.sh
+++ b/src/sys/tests/scripts/03_wait_ready.sh
@@ -27,25 +27,34 @@ wait_http() {

 echo "[INFO] Waiting for ES/Kibana/Master/Fluent Bit/Bind..."

+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
 # ES (>= yellow)
 attempt=1; max=120
+ES_T0=$(date +%s)
 while (( attempt <= max )); do
-  if curl -fsS "http://localhost:9200/_cluster/health?wait_for_status=yellow&timeout=1s" >/dev/null 2>&1; then
+  if curl -fsS "http://localhost:${ES_HTTP_PORT:-9200}/_cluster/health?wait_for_status=yellow&timeout=1s" >/dev/null 2>&1; then
    break
  fi
  echo "[..] waiting ES ($attempt/$max)"; sleep 5; ((attempt++))
 done
 [[ $attempt -le $max ]] || { echo "[ERR] ES not ready" >&2; exit 1; }
+ES_T1=$(date +%s); echo "[TIME] ES ready in $((ES_T1-ES_T0))s"

 # Kibana: must be HTTP 200 and overall.level=available
 echo "[INFO] Waiting for Kibana to be available (HTTP 200)..."
 kb_attempt=1; kb_max=180
+KB_T0=$(date +%s)
 while (( kb_attempt <= kb_max )); do
-  body=$(curl -sS "http://localhost:5601/api/status" 2>/dev/null || true)
-  code=$(curl -s -o /dev/null -w "%{http_code}" "http://localhost:5601/api/status" || echo 000)
+  body=$(curl -sS "http://localhost:${KIBANA_PORT:-5601}/api/status" 2>/dev/null || true)
+  code=$(curl -s -o /dev/null -w "%{http_code}" "http://localhost:${KIBANA_PORT:-5601}/api/status" || echo 000)
  if [[ "$code" == "200" ]]; then
    if echo "$body" | grep -q '"level":"available"'; then
-      echo "[OK] Kibana available (HTTP 200)"
+      KB_T1=$(date +%s)
+      echo "[OK] Kibana available (HTTP 200) in $((KB_T1-KB_T0))s"
      break
    fi
  fi
@@ -58,11 +67,13 @@ if (( kb_attempt > kb_max )); then
 fi

 # Master
-wait_http "http://localhost:32300/readyz" 120
+MASTER_T0=$(date +%s)
+wait_http "http://localhost:${MASTER_PORT:-32300}/readyz" 120
+MASTER_T1=$(date +%s); echo "[TIME] Master readyz in $((MASTER_T1-MASTER_T0))s"

 # Fluent Bit (host metrics on host ports)
-wait_http "http://localhost:2020/api/v2/metrics" 120
-wait_http "http://localhost:2021/api/v2/metrics" 120
+FB1_T0=$(date +%s); wait_http "http://localhost:${NODE_A_PORT:-2020}/api/v2/metrics" 120; FB1_T1=$(date +%s); echo "[TIME] FluentBit:${NODE_A_PORT:-2020} in $((FB1_T1-FB1_T0))s"
+FB2_T0=$(date +%s); wait_http "http://localhost:${NODE_B_PORT:-2021}/api/v2/metrics" 120; FB2_T1=$(date +%s); echo "[TIME] FluentBit:${NODE_B_PORT:-2021} in $((FB2_T1-FB2_T0))s"

 # Bind config check
 BIND_ID="$(service_id bind)"
@@ -72,4 +83,63 @@ else
  echo "[WARN] bind container id not found"
 fi

+# ========== Additional module readiness checks ==========
+
+# Prometheus
+PROM_T0=$(date +%s); wait_http "http://localhost:${PROMETHEUS_PORT:-9090}/-/ready" 120; PROM_T1=$(date +%s); echo "[TIME] Prometheus ready in $((PROM_T1-PROM_T0))s"
+
+# Grafana health (database: ok)
+echo "[INFO] Waiting for Grafana health..."
+gf_attempt=1; gf_max=120
+while (( gf_attempt <= gf_max )); do
+  gf_body=$(curl -sS "http://localhost:${GRAFANA_PORT:-3000}/api/health" 2>/dev/null || true)
+  gf_code=$(curl -s -o /dev/null -w "%{http_code}" "http://localhost:${GRAFANA_PORT:-3000}/api/health" || echo 000)
+  if [[ "$gf_code" == "200" ]] && echo "$gf_body" | grep -q '"database"\s*:\s*"ok"'; then
+    echo "[OK] Grafana health database=ok"
+    break
+  fi
+  echo "[..] waiting grafana health ($gf_attempt/$gf_max), last_code=$gf_code"
+  sleep 3; ((gf_attempt++))
+done
+if (( gf_attempt > gf_max )); then
+  echo "[ERR] Grafana /api/health not ready" >&2; exit 1
+fi
+
+# Alertmanager
+wait_http "http://localhost:${ALERTMANAGER_PORT:-9093}/api/v2/status" 120
+
+# Web proxy checks（按端口细化）
+code_for() { curl -s -o /dev/null -w "%{http_code}" "$1" || echo 000; }
+header_val() { curl -s -D - -o /dev/null "$@" | awk -F': ' 'BEGIN{IGNORECASE=1}$1=="Access-Control-Allow-Origin"{gsub("\r","",$2);print $2}'; }
+
+echo "[INFO] Checking web-proxy ports..."
+
+# 8080 首页必须 200
+tries=1; max=60; P8080_T0=$(date +%s)
+while (( tries <= max )); do
+  c=$(code_for "http://localhost:${WEB_PROXY_PORT_8080:-8080}/")
+  if [[ "$c" == "200" ]]; then P8080_T1=$(date +%s); echo "[OK] 8080 / ($c) in $((P8080_T1-P8080_T0))s"; break; fi
+  echo "[..] waiting 8080/ ($tries/$max), code=$c"; sleep 3; ((tries++))
+done
+(( tries <= max )) || { echo "[ERR] 8080/ not ready" >&2; exit 1; }
+
+# 8083 Kibana 允许 200/302（上面已就绪，端口侧再快速确认）
+tries=1; max=40; P8083_T0=$(date +%s)
+while (( tries <= max )); do
+  c=$(code_for "http://localhost:${WEB_PROXY_PORT_8083:-8083}/")
+  if [[ "$c" == "200" || "$c" == "302" ]]; then P8083_T1=$(date +%s); echo "[OK] 8083 / ($c) in $((P8083_T1-P8083_T0))s"; break; fi
+  echo "[..] waiting 8083/ ($tries/$max), code=$c"; sleep 3; ((tries++))
+done
+(( tries <= max )) || { echo "[ERR] 8083/ not ready" >&2; exit 1; }
+
+# 8084 Alertmanager + CORS
+P8084_T0=$(date +%s); wait_http "http://localhost:${WEB_PROXY_PORT_8084:-8084}/api/v2/status" 60; P8084_T1=$(date +%s)
+cors=$(header_val -H "Origin: http://localhost:${WEB_PROXY_PORT_8080:-8080}" "http://localhost:${WEB_PROXY_PORT_8084:-8084}/api/v2/status" || true)
+if [[ -z "$cors" ]]; then echo "[ERR] 8084 CORS missing" >&2; exit 1; else echo "[OK] 8084 CORS: $cors in $((P8084_T1-P8084_T0))s"; fi
+
+# 8085 Master /readyz + CORS（API 走 8085 才需跨域）
+P8085_T0=$(date +%s); wait_http "http://localhost:${WEB_PROXY_PORT_8085:-8085}/readyz" 60; P8085_T1=$(date +%s)
+cors=$(header_val -H "Origin: http://localhost:${WEB_PROXY_PORT_8080:-8080}" "http://localhost:${WEB_PROXY_PORT_8085:-8085}/api/v1/master/nodes" || true)
+if [[ -z "$cors" ]]; then echo "[ERR] 8085 CORS missing" >&2; exit 1; else echo "[OK] 8085 CORS: $cors in $((P8085_T1-P8085_T0))s"; fi
+
 echo "[OK] All services are ready"
--- a/src/sys/tests/scripts/04_verify_dns_routing.sh
+++ b/src/sys/tests/scripts/04_verify_dns_routing.sh
@@ -4,20 +4,15 @@ set -euo pipefail
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"

-compose() {
-  if docker compose version >/dev/null 2>&1; then
-    docker compose "$@"
-  else
-    docker-compose "$@"
-  fi
-}
-
-service_id() {
-  compose -p argus-sys ps -q "$1"
+# 直接根据 container_name 获取容器ID，避免 compose project 名称不一致导致查找失败
+cid_by_name() {
+  docker ps -aqf "name=^$1$"
 }

 echo "[INFO] Verifying DNS routing via bind..."

+pushd "$TEST_ROOT" >/dev/null
+
 # Check master IP file exists in shared private
 MASTER_FILE="$TEST_ROOT/private/argus/etc/master.argus.com"
 if [[ ! -f "$MASTER_FILE" ]]; then
@@ -28,7 +23,7 @@ MASTER_IP_HOST="$(cat "$MASTER_FILE" | tr -d '\r\n' || true)"
 echo "[INFO] master.argus.com file content: ${MASTER_IP_HOST}"

 # dig inside bind container
-BIN_ID="$(service_id bind)"
+BIN_ID="$(cid_by_name argus-bind-sys)"
 if [[ -n "$BIN_ID" ]]; then
  DIG_IP="$(docker exec "$BIN_ID" dig +short master.argus.com A | tail -n1 || true)"
  echo "[INFO] dig(master.argus.com) from bind container -> $DIG_IP"
@@ -39,8 +34,27 @@ else
  echo "[WARN] bind container not found; skip dig"
 fi

-for node in node-a node-b; do
-  CID="$(service_id "$node")"
+check_inside() {
+  local cname="$1"; shift
+  local domains=("$@")
+  CID="$(cid_by_name "$cname")"
+  if [[ -z "$CID" ]]; then
+    echo "[WARN] container $cname not found; skip"
+    return 0
+  fi
+  for d in "${domains[@]}"; do
+    echo "[INFO] Checking resolution inside $cname for $d..."
+    if ! docker exec "$CID" getent hosts "$d" >/dev/null 2>&1; then
+      echo "[ERR] $cname cannot resolve $d" >&2
+      return 1
+    fi
+    RES="$(docker exec "$CID" getent hosts "$d" | awk '{print $1}' | head -n1)"
+    echo "[OK] $cname resolved $d -> $RES"
+  done
+}
+
+for node in argus-node-a argus-node-b; do
+  CID="$(cid_by_name "$node")"
  echo "[INFO] Checking resolution inside $node..."
  if ! docker exec "$CID" getent hosts master.argus.com >/dev/null 2>&1; then
    echo "[ERR] $node cannot resolve master.argus.com" >&2
@@ -50,5 +64,10 @@ for node in node-a node-b; do
  echo "[OK] $node resolved master.argus.com -> $RES"
 done

-echo "[OK] DNS routing verified"
+popd >/dev/null

+# 追加：在 metric 节点中验证 master 与 prom 域名解析
+check_inside argus-metric-test-node master.argus.com prom.metric.argus.com || exit 1
+check_inside argus-metric-test-gpu-node master.argus.com prom.metric.argus.com || exit 1
+
+echo "[OK] DNS routing verified"
--- a/src/sys/tests/scripts/05_agent_register.sh
+++ b/src/sys/tests/scripts/05_agent_register.sh
@@ -5,7 +5,12 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
 TMP_DIR="$TEST_ROOT/tmp"

-API_BASE="http://localhost:32300/api/v1/master"
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+API_BASE="http://localhost:${MASTER_PORT:-32300}/api/v1/master"

 HOST_A="dev-yyrshare-nbnyx10-cp2f-pod-0"
 HOST_B="dev-yyrshare-uuuu10-ep2f-pod-0"
@@ -49,8 +54,35 @@ for _ in {1..60}; do
  fi
 done

+# 若仍未全部注册，尝试重启 node-b 并再等待一轮（兼容 DNS/启动时序抖动）
 if [[ ! -s "$TMP_DIR/node_id_a" || ! -s "$TMP_DIR/node_id_b" ]]; then
-  echo "[ERR] Agents did not register in time" >&2
+  echo "[WARN] node-a or node-b not registered in first window; restarting node-b and retrying..." >&2
+  # 仅重启 node-b，避免影响 es/kibana/master
+  if docker ps --format '{{.Names}}' | grep -q '^argus-node-b$'; then
+    docker restart argus-node-b >/dev/null 2>&1 || true
+  fi
+  # 再等待一轮（最多 120 秒）
+  > "$TMP_DIR/node_id_b"
+  for _ in {1..60}; do
+    sleep 2
+    resp=$(curl -fsS "$API_BASE/nodes" 2>/dev/null || true)
+    [[ -z "$resp" ]] && continue
+    if ! echo "$resp" | head -c1 | grep -q '\['; then
+      continue
+    fi
+    echo "$resp" > "$TMP_DIR/nodes_list.json"
+    ID_A=$(extract_node "$HOST_A" "$TMP_DIR/node_id_a" "$TMP_DIR/nodes_list.json" 2>/dev/null || true)
+    ID_B=$(extract_node "$HOST_B" "$TMP_DIR/node_id_b" "$TMP_DIR/nodes_list.json" 2>/dev/null || true)
+    if [[ -s "$TMP_DIR/node_id_a" && -s "$TMP_DIR/node_id_b" ]]; then
+      break
+    fi
+  done
+fi
+
+if [[ ! -s "$TMP_DIR/node_id_a" || ! -s "$TMP_DIR/node_id_b" ]]; then
+  echo "[ERR] Agents did not register in time (after retry)" >&2
+  echo "[HINT] Current /nodes response:" >&2
+  sed -n '1,200p' "$TMP_DIR/nodes_list.json" >&2 || true
  exit 1
 fi

--- a/src/sys/tests/scripts/06_write_health_and_assert.sh
+++ b/src/sys/tests/scripts/06_write_health_and_assert.sh
@@ -5,7 +5,12 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
 TMP_DIR="$TEST_ROOT/tmp"

-API_BASE="http://localhost:32300/api/v1/master"
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+API_BASE="http://localhost:${MASTER_PORT:-32300}/api/v1/master"

 HOST_A="dev-yyrshare-nbnyx10-cp2f-pod-0"
 HOST_B="dev-yyrshare-uuuu10-ep2f-pod-0"
--- a/src/sys/tests/scripts/07_logs_send_and_assert.sh
+++ b/src/sys/tests/scripts/07_logs_send_and_assert.sh
@@ -3,9 +3,25 @@ set -euo pipefail

 echo "[INFO] Sending logs via node-a/node-b and asserting ES counts..."

+# 载入端口变量
+TEST_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")"/.. && pwd)"
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+# Robust count helper: tolerates 404/503 and non-JSON responses, returns integer >=0
 get_count() {
-  local idx="$1"
-  curl -s "http://localhost:9200/${idx}/_count?ignore_unavailable=true&allow_no_indices=true" | sed -E 's/.*"count":([0-9]+).*/\1/' | awk 'NF{print $0;exit} END{if(NR==0)print 0}'
+  local idx="$1"; local tmp; tmp=$(mktemp)
+  local code
+  code=$(curl -s -o "$tmp" -w "%{http_code}" "http://localhost:${ES_HTTP_PORT:-9200}/${idx}/_count?ignore_unavailable=true&allow_no_indices=true" || true)
+  if [[ "$code" == "200" ]]; then
+    local val
+    val=$(jq -r '(.count // 0) | tonumber? // 0' "$tmp" 2>/dev/null || echo 0)
+    echo "$val"
+  else
+    echo 0
+  fi
+  rm -f "$tmp"
 }

 train0=$(get_count "train-*")
@@ -32,11 +48,26 @@ send_logs "$node_a" "host01"
 send_logs "$node_b" "host02"

 echo "[INFO] Waiting for ES to ingest..."
-sleep 10
+# Proactively refresh indices (ignore errors if not created yet)
+curl -s -X POST "http://localhost:${ES_HTTP_PORT:-9200}/train-*/_refresh" >/dev/null 2>&1 || true
+curl -s -X POST "http://localhost:${ES_HTTP_PORT:-9200}/infer-*/_refresh" >/dev/null 2>&1 || true

-train1=$(get_count "train-*")
-infer1=$(get_count "infer-*")
-final=$((train1 + infer1))
+# Retry up to 120s for counts to increase and reach threshold (>=4)
+final=0
+threshold=4
+for attempt in {1..60}; do
+  train1=$(get_count "train-*")
+  infer1=$(get_count "infer-*")
+  final=$((train1 + infer1))
+  if (( final > base && final >= threshold )); then
+    break
+  fi
+  echo "[..] waiting ES counts increase to >=${threshold} ($attempt/60) current=${final} base=${base}"
+  # refresh indices again to speed up visibility
+  curl -s -X POST "http://localhost:${ES_HTTP_PORT:-9200}/train-*/_refresh" >/dev/null 2>&1 || true
+  curl -s -X POST "http://localhost:${ES_HTTP_PORT:-9200}/infer-*/_refresh" >/dev/null 2>&1 || true
+  sleep 2
+done
 echo "[INFO] final counts: train=${train1} infer=${infer1} total=${final}"

 if (( final <= base )); then
@@ -44,19 +75,20 @@ if (( final <= base )); then
  exit 1
 fi

+# Minimal threshold to be tolerant: expect at least 4 documents (2 train + 1 infer per node)
 if (( final < 4 )); then
  echo "[ERR] ES total below expected threshold: ${final} < 4" >&2
  exit 1
 fi

 # Health endpoints
-es_health=$(curl -s "http://localhost:9200/_cluster/health" | grep -o '"status":"[^"]*"' | cut -d'"' -f4)
+es_health=$(curl -s "http://localhost:${ES_HTTP_PORT:-9200}/_cluster/health" | grep -o '"status":"[^"]*"' | cut -d'"' -f4)
 if [[ "$es_health" != "green" && "$es_health" != "yellow" ]]; then
  echo "[ERR] ES health not green/yellow: $es_health" >&2
  exit 1
 fi

-if ! curl -fs "http://localhost:5601/api/status" >/dev/null 2>&1; then
+if ! curl -fs "http://localhost:${KIBANA_PORT:-5601}/api/status" >/dev/null 2>&1; then
  echo "[WARN] Kibana status endpoint not available"
 fi

--- a/src/sys/tests/scripts/08_restart_agent_reregister.sh
+++ b/src/sys/tests/scripts/08_restart_agent_reregister.sh
@@ -6,7 +6,12 @@ TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
 TMP_DIR="$TEST_ROOT/tmp"
 REPO_ROOT="$(cd "$TEST_ROOT/../../.." && pwd)"

-API_BASE="http://localhost:32300/api/v1/master"
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+API_BASE="http://localhost:${MASTER_PORT:-32300}/api/v1/master"

 if [[ -f "$TEST_ROOT/.env" ]]; then
  set -a
@@ -49,7 +54,7 @@ compose() {
  fi
 }

-echo "[INFO] Recreating node-b with static IP 172.29.0.200..."
+echo "[INFO] Recreating node-b with static IP 172.31.0.200..."
 pushd "$TEST_ROOT" >/dev/null
 compose -p argus-sys rm -sf node-b || true
 popd >/dev/null
@@ -58,19 +63,34 @@ docker rm -f argus-node-b >/dev/null 2>&1 || true

 AGENT_BIN_PATH="$(cat "$TMP_DIR/agent_binary_path")"

+# 选择 compose 管理的网络名（默认 argus-sys_sysnet）。
+detect_sysnet() {
+  if docker network inspect argus-sys_sysnet >/dev/null 2>&1; then
+    echo argus-sys_sysnet; return
+  fi
+  # 回退：从 master 容器推断所连网络（取第一个）
+  local n
+  n=$(docker inspect -f '{{range $k, $_ := .NetworkSettings.Networks}}{{println $k}}{{end}}' argus-master-sys 2>/dev/null | head -n1 || true)
+  if [[ -n "$n" ]]; then echo "$n"; return; fi
+  # 最后兜底：尝试项目默认网络（不保证有 IPAM）
+  echo argus-sys_default
+}
+SYSNET_NAME=$(detect_sysnet)
+echo "[INFO] Using docker network: $SYSNET_NAME"
+
 docker run -d \
  --name argus-node-b \
  --hostname dev-yyrshare-uuuu10-ep2f-pod-0 \
-  --network argus-sys-net \
-  --ip 172.29.0.200 \
-  --dns 172.29.0.2 \
+  --network "$SYSNET_NAME" \
+  --ip 172.31.0.200 \
+  --dns 172.31.0.2 \
  -e MASTER_ENDPOINT=http://master.argus.com:3000 \
  -e REPORT_INTERVAL_SECONDS=2 \
  -e ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133} \
  -e ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015} \
  -e ES_HOST=es \
  -e ES_PORT=9200 \
-  -p 2021:2020 \
+  -p ${NODE_B_PORT:-2021}:2020 \
  -v "$TEST_ROOT/private-nodeb/argus/agent/dev-yyrshare-uuuu10-ep2f-pod-0:/private/argus/agent/dev-yyrshare-uuuu10-ep2f-pod-0" \
  -v "$AGENT_BIN_PATH:/usr/local/bin/argus-agent:ro" \
  -v "$SCRIPT_DIR/node_entrypoint.sh:/usr/local/bin/node-entrypoint.sh:ro" \
@@ -90,15 +110,15 @@ node=json.load(open(sys.argv[1]))
 last0=sys.argv[2]
 ip=node.get("meta_data",{}).get("ip")
 lu=node.get("last_updated")
-assert ip=="172.29.0.200"
+assert ip=="172.31.0.200"
 assert lu and lu!=last0
 PY
    then
-      echo "[OK] node-b re-registered with new IP 172.29.0.200"
+      echo "[OK] node-b re-registered with new IP 172.31.0.200"
      exit 0
    fi
  fi
 done

-echo "[ERR] node-b did not update to IP 172.29.0.200 in time" >&2
+echo "[ERR] node-b did not update to IP 172.31.0.200 in time" >&2
 exit 1
--- a/src/sys/tests/scripts/09_down.sh
+++ b/src/sys/tests/scripts/09_down.sh
@@ -12,12 +12,33 @@ compose() {
  fi
 }

-docker rm -f argus-node-b >/dev/null 2>&1 || true
-
 pushd "$TEST_ROOT" >/dev/null
 compose -p argus-sys down --remove-orphans || true
+compose down --remove-orphans || true
 popd >/dev/null

+echo "[INFO] Force removing containers by name (if any)..."
+containers=(
+  argus-node-a
+  argus-node-b
+  argus-metric-test-node
+  argus-grafana
+  argus-kibana-sys
+  argus-master-sys
+  argus-bind-sys
+  argus-ftp
+  argus-es-sys
+  argus-prometheus
+)
+for c in "${containers[@]}"; do
+  id=$(docker ps -aqf "name=^${c}$" || true)
+  if [[ -n "$id" ]]; then
+    docker rm -f "$id" >/dev/null 2>&1 || true
+  fi
+done
+
+echo "[INFO] Removing compose networks (handled by compose down)"
+
 echo "[INFO] Cleaning private directories..."
 if [[ -d "$TEST_ROOT/private" ]]; then
  docker run --rm -v "$TEST_ROOT/private:/target" ubuntu:24.04 chown -R "$(id -u):$(id -g)" /target >/dev/null 2>&1 || true
--- a/src/sys/tests/scripts/10_metric_publish.sh
+++ b/src/sys/tests/scripts/10_metric_publish.sh
@@ -0,0 +1,89 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+REPO_ROOT="$(cd "$TEST_ROOT/../../.." && pwd)"
+
+PLUGIN_DIR="$REPO_ROOT/src/metric/client-plugins/all-in-one-full"
+FTP_CONTAINER="argus-ftp"
+
+if [[ ! -d "$PLUGIN_DIR" ]]; then
+  echo "[SYS-METRIC] Metric client plugin directory not found: $PLUGIN_DIR" >&2
+  exit 1
+fi
+
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  # shellcheck source=/dev/null
+  source "$TEST_ROOT/.env"
+fi
+
+OWNER="${ARGUS_BUILD_UID:-2133}:${ARGUS_BUILD_GID:-2015}"
+
+resolve_output_dir() {
+  local host_mount
+  if docker ps --format '{{.Names}}' | grep -q "^${FTP_CONTAINER}$"; then
+    host_mount=$(docker inspect "$FTP_CONTAINER" --format '{{range .Mounts}}{{if eq .Destination "/private/argus/ftp"}}{{.Source}}{{end}}{{end}}' 2>/dev/null || true)
+    if [[ -n "$host_mount" ]]; then
+      echo "$host_mount/share"
+      return 0
+    fi
+  fi
+  echo "$TEST_ROOT/private/argus/metric/ftp/share"
+}
+
+OUTPUT_DIR="$(resolve_output_dir)"
+mkdir -p "$OUTPUT_DIR"
+
+if [[ ! -w "$OUTPUT_DIR" ]]; then
+  echo "[SYS-METRIC] 无法写入 FTP 输出目录: $OUTPUT_DIR" >&2
+  echo "             请确认目录权限与 ARGUS_BUILD_UID/GID 一致" >&2
+  exit 1
+fi
+
+pushd "$PLUGIN_DIR" >/dev/null
+
+# --- Inject agent binary built in 01_bootstrap (if present) ---
+AGENT_PATH_FILE="$TEST_ROOT/tmp/agent_binary_path"
+AGENT_BIN_CANDIDATE="$REPO_ROOT/src/agent/dist/argus-agent"
+if [[ -f "$AGENT_PATH_FILE" ]]; then
+  AGENT_BIN="$(tr -d '\n' < "$AGENT_PATH_FILE")"
+else
+  AGENT_BIN="$AGENT_BIN_CANDIDATE"
+fi
+
+if [[ -x "$AGENT_BIN" ]]; then
+  echo "[SYS-METRIC] 使用 01 阶段构建的 agent: $AGENT_BIN"
+  TARGET_BIN="plugins/argus-agent/bin/argus-agent"
+  if [[ -f "$TARGET_BIN" ]]; then
+    cp -f "$AGENT_BIN" "$TARGET_BIN"
+  else
+    mkdir -p "$(dirname "$TARGET_BIN")"
+    cp "$AGENT_BIN" "$TARGET_BIN"
+  fi
+  chmod +x "$TARGET_BIN"
+else
+  echo "[SYS-METRIC] 未找到可执行的 agent 二进制（预期: $AGENT_BIN），继续使用插件目录内置版本"
+fi
+
+echo "[SYS-METRIC] Bumping metric artifact version..."
+bash scripts/version-manager.sh bump minor
+
+VERSION_FILE="config/VERSION"
+if [[ ! -f "$VERSION_FILE" ]]; then
+  echo "[SYS-METRIC] VERSION 文件缺失: $VERSION_FILE" >&2
+  exit 1
+fi
+
+VERSION=$(tr -d '\n' < "$VERSION_FILE")
+echo "[SYS-METRIC] 当前版本: $VERSION"
+
+echo "[SYS-METRIC] Packaging metric artifact..."
+bash scripts/package_artifact.sh --force
+
+echo "[SYS-METRIC] Publishing artifact to FTP share..."
+bash scripts/publish_artifact.sh "$VERSION" --output-dir "$OUTPUT_DIR" --owner "$OWNER"
+
+popd >/dev/null
+
+echo "[SYS-METRIC] Metric artifact published to $OUTPUT_DIR"
--- a/src/sys/tests/scripts/11_metric_node_install.sh
+++ b/src/sys/tests/scripts/11_metric_node_install.sh
@@ -0,0 +1,50 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  # shellcheck source=/dev/null
+  source "$TEST_ROOT/.env"
+fi
+
+CONTAINER="argus-metric-test-node"
+
+if ! docker ps --format '{{.Names}}' | grep -q "^${CONTAINER}$"; then
+  echo "[SYS-METRIC] 容器 ${CONTAINER} 未运行，无法执行安装" >&2
+  exit 1
+fi
+
+FTP_HOST="${FTP_SERVER:-172.31.0.40}"
+FTP_USER="${FTP_USER:-ftpuser}"
+FTP_PASSWORD="${FTP_PASSWORD:-ZGClab1234!}"
+FTP_PORT="${FTP_PORT:-21}"
+
+echo "[SYS-METRIC] 在 ${CONTAINER} 内执行安装 (FTP: ${FTP_HOST}:${FTP_PORT})"
+
+docker exec \
+  -e FTP_HOST="$FTP_HOST" \
+  -e FTP_USER="$FTP_USER" \
+  -e FTP_PASSWORD="$FTP_PASSWORD" \
+  -e FTP_PORT="$FTP_PORT" \
+  "$CONTAINER" bash -c '
+set -e
+
+if ! command -v curl &>/dev/null; then
+  echo "[SYS-METRIC] curl 未安装，开始安装依赖..."
+  apt-get update >/dev/null && apt-get install -y curl >/dev/null
+fi
+
+cd /tmp
+echo "[SYS-METRIC] 下载 setup.sh..."
+curl -u "${FTP_USER}:${FTP_PASSWORD}" "ftp://${FTP_HOST}:${FTP_PORT}/setup.sh" -o setup.sh
+
+echo "[SYS-METRIC] 执行安装..."
+chmod +x setup.sh
+bash setup.sh --server "${FTP_HOST}" --user "${FTP_USER}" --password "${FTP_PASSWORD}" --port "${FTP_PORT}"
+
+echo "[SYS-METRIC] 安装完成"
+'
+
+echo "[SYS-METRIC] Metric test node 安装流程完成"
--- a/src/sys/tests/scripts/12_metric_gpu_install.sh
+++ b/src/sys/tests/scripts/12_metric_gpu_install.sh
@@ -0,0 +1,82 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+
+ENABLE_GPU=${ARGUS_SYS_ENABLE_GPU:-false}
+
+if [[ "$ENABLE_GPU" != "true" ]]; then
+  echo "[SYS-METRIC] 未启用 GPU 流程，跳过 GPU 节点安装"
+  exit 0
+fi
+
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  # shellcheck source=/dev/null
+  source "$TEST_ROOT/.env"
+fi
+
+CONTAINER="argus-metric-test-gpu-node"
+
+if ! docker ps --format '{{.Names}}' | grep -q "^${CONTAINER}$"; then
+  echo "[SYS-METRIC] 预期启动的 ${CONTAINER} 未运行" >&2
+  exit 1
+fi
+
+FTP_HOST="${FTP_SERVER:-172.31.0.40}"
+FTP_USER="${FTP_USER:-ftpuser}"
+FTP_PASSWORD="${FTP_PASSWORD:-ZGClab1234!}"
+FTP_PORT="${FTP_PORT:-21}"
+
+echo "[SYS-METRIC] 在 GPU 节点执行安装 (FTP: ${FTP_HOST}:${FTP_PORT})"
+
+docker exec \
+  -e FTP_HOST="$FTP_HOST" \
+  -e FTP_USER="$FTP_USER" \
+  -e FTP_PASSWORD="$FTP_PASSWORD" \
+  -e FTP_PORT="$FTP_PORT" \
+  "$CONTAINER" bash -c '
+set -e
+
+if ! command -v nvidia-smi &>/dev/null; then
+  echo "[SYS-METRIC] GPU 节点缺少 nvidia-smi" >&2
+  exit 1
+fi
+
+nvidia-smi >/dev/null || true
+
+if ! command -v curl &>/dev/null; then
+  echo "[SYS-METRIC] curl 未安装，开始安装依赖..."
+  apt-get update >/dev/null && apt-get install -y curl >/dev/null
+fi
+
+cd /tmp
+echo "[SYS-METRIC] 下载 setup.sh..."
+curl -u "${FTP_USER}:${FTP_PASSWORD}" "ftp://${FTP_HOST}:${FTP_PORT}/setup.sh" -o setup.sh
+
+echo "[SYS-METRIC] 执行安装..."
+chmod +x setup.sh
+bash setup.sh --server "${FTP_HOST}" --user "${FTP_USER}" --password "${FTP_PASSWORD}" --port "${FTP_PORT}"
+
+echo "[SYS-METRIC] GPU 节点安装完成"
+'
+
+echo "[SYS-METRIC] Metric GPU 节点安装流程完成"
+
+# 就绪性检测：9400(dcgm) 与 9100(node) 端口
+echo "[SYS-METRIC] 等待 dcgm-exporter(9400) 与 node-exporter(9100) 就绪..."
+retries=30
+until docker exec "$CONTAINER" bash -lc "curl -fsS --max-time 2 http://localhost:9400/metrics >/dev/null"; do
+  ((retries--)) || { echo "[ERR] dcgm-exporter 9400 未就绪" >&2; exit 1; }
+  sleep 2
+done
+echo "[OK] dcgm-exporter 端点可访问"
+
+retries=30
+until docker exec "$CONTAINER" bash -lc "curl -fsS --max-time 2 http://localhost:9100/metrics >/dev/null"; do
+  ((retries--)) || { echo "[ERR] node-exporter 9100 未就绪" >&2; exit 1; }
+  sleep 2
+done
+echo "[OK] node-exporter 端点可访问"
+
+mkdir -p "$TEST_ROOT/tmp" && touch "$TEST_ROOT/tmp/gpu_install_ready"
--- a/src/sys/tests/scripts/13_metric_verify.sh
+++ b/src/sys/tests/scripts/13_metric_verify.sh
@@ -0,0 +1,40 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+echo "[SYS-METRIC] Verify: master"
+"$SCRIPT_DIR/13_metric_verify_master.sh"
+echo
+
+echo "[SYS-METRIC] Verify: prometheus"
+PROM_RETRIES=${PROM_VERIFY_RETRIES:-2}
+PROM_BACKOFF=${PROM_VERIFY_BACKOFF_SECONDS:-30}
+attempt=0
+while true; do
+  if "$SCRIPT_DIR/13_metric_verify_prometheus.sh"; then
+    break
+  fi
+  attempt=$((attempt+1))
+  if (( attempt > PROM_RETRIES )); then
+    echo "[ERR] prometheus verify failed after $PROM_RETRIES retries" >&2
+    exit 1
+  fi
+  echo "[WARN] prometheus verify failed; retry $attempt/$PROM_RETRIES after ${PROM_BACKOFF}s"
+  sleep "$PROM_BACKOFF"
+done
+echo
+
+echo "[SYS-METRIC] Verify: dataplane"
+"$SCRIPT_DIR/13_metric_verify_dataplane.sh"
+echo
+
+echo "[SYS-METRIC] Verify: grafana"
+"$SCRIPT_DIR/13_metric_verify_grafana.sh"
+echo
+
+echo "[SYS-METRIC] Verify: grafana panels"
+"$SCRIPT_DIR/13_metric_verify_grafana_panels.sh"
+echo
+
+echo "[SYS-METRIC] Metric verification completed"
--- a/src/sys/tests/scripts/13_metric_verify_dataplane.sh
+++ b/src/sys/tests/scripts/13_metric_verify_dataplane.sh
@@ -0,0 +1,66 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+TMP_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")"/.. && pwd)/tmp/metric-verify"
+mkdir -p "$TMP_DIR"
+
+# 载入端口变量（由 .env 提供）
+TEST_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")"/.. && pwd)"
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+PROM_BASE="http://localhost:${PROMETHEUS_PORT:-9090}/api/v1"
+INSTANCE="${METRIC_TEST_INSTANCE:-172.31.0.50:9100}"
+IP_ONLY="${INSTANCE%%:*}"
+
+echo "[VERIFY:DATA] node exporter metrics present in container"
+docker exec argus-metric-test-node bash -lc "curl -fsS --max-time 5 http://localhost:9100/metrics | head -n 5" > "$TMP_DIR/node_metrics_head.txt" || { echo "[ERR] cannot fetch node exporter metrics" >&2; exit 1; }
+if ! grep -E "node_(exporter_build_info|time_seconds)" -q "$TMP_DIR/node_metrics_head.txt"; then
+  echo "[WARN] head did not show expected lines; continuing (exporter may output later lines)"
+fi
+echo "[OK] node exporter endpoint reachable"
+
+echo "[VERIFY:DATA] Prometheus has recent sample for build_info"
+curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=node_exporter_build_info{job=\"node\",ip=\"$IP_ONLY\"}" > "$TMP_DIR/prom_ne_build_info_1.json"
+
+python3 - "$TMP_DIR/prom_ne_build_info_1.json" <<'PY'
+import json,sys,time
+j=json.load(open(sys.argv[1]))
+res=j.get('data',{}).get('result',[])
+assert res, 'no result for node_exporter_build_info'
+ts=float(res[0]['value'][0])
+now=time.time()
+assert now-ts<180, f"sample too old: now={now} ts={ts}"
+print(int(ts))
+PY
+T1=$?
+sleep 30
+curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=node_exporter_build_info{job=\"node\",ip=\"$IP_ONLY\"}" > "$TMP_DIR/prom_ne_build_info_2.json"
+
+TS1=$(python3 - "$TMP_DIR/prom_ne_build_info_1.json" <<'PY'
+import json,sys
+print(float(json.load(open(sys.argv[1]))['data']['result'][0]['value'][0]))
+PY
+)
+TS2=$(python3 - "$TMP_DIR/prom_ne_build_info_2.json" <<'PY'
+import json,sys
+print(float(json.load(open(sys.argv[1]))['data']['result'][0]['value'][0]))
+PY
+)
+awk -v a="$TS1" -v b="$TS2" 'BEGIN{ if (b>=a) exit 0; else exit 1 }' || { echo "[ERR] sample timestamp did not advance" >&2; exit 1; }
+echo "[OK] sample timestamp advanced"
+echo "[DONE] dataplane verify"
+
+# 追加：GPU 节点端点连通性检查（启用 GPU 时）
+if [[ "${ENABLE_GPU:-false}" == "true" ]]; then
+  echo
+  echo "[VERIFY:DATA][GPU] curl endpoints on gpu node"
+  if ! docker exec argus-metric-test-gpu-node bash -lc 'curl -fsS --max-time 5 http://localhost:9100/metrics >/dev/null'; then
+    echo "[ERR] gpu node 9100 not reachable" >&2; exit 1
+  fi
+  if ! docker exec argus-metric-test-gpu-node bash -lc 'curl -fsS --max-time 5 http://localhost:9400/metrics >/dev/null'; then
+    echo "[ERR] gpu node 9400 not reachable" >&2; exit 1
+  fi
+  echo "[OK] gpu node endpoints reachable"
+fi
--- a/src/sys/tests/scripts/13_metric_verify_grafana.sh
+++ b/src/sys/tests/scripts/13_metric_verify_grafana.sh
@@ -0,0 +1,44 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+TEST_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")"/.. && pwd)"
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+PROM_DOMAIN="prom.metric.argus.com:${PROMETHEUS_PORT:-9090}"
+GRAF="http://localhost:${GRAFANA_PORT:-3000}"
+
+echo "[VERIFY:GRAFANA] /api/health"
+TMP_FILE="$(cd "$(dirname "$0")"/.. && pwd)/tmp/metric-verify/graf_health.json"
+mkdir -p "$(dirname "$TMP_FILE")"
+curl -fsS --max-time 10 "$GRAF/api/health" -o "$TMP_FILE" || { echo "[ERR] failed to GET /api/health" >&2; exit 1; }
+python3 - "$TMP_FILE" <<'PY'
+import sys,json
+with open(sys.argv[1],'r',encoding='utf-8') as f:
+    j=json.load(f)
+assert j.get('database')=='ok', f"health not ok: {j}"
+print('OK')
+PY
+
+echo "[VERIFY:GRAFANA] datasource URL uses domain: $PROM_DOMAIN"
+DS_FILE="/private/argus/metric/grafana/provisioning/datasources/datasources.yml"
+if ! docker exec argus-grafana sh -lc "test -f $DS_FILE"; then
+  DS_FILE="/etc/grafana/provisioning/datasources/datasources.yml"
+fi
+docker exec argus-grafana sh -lc "grep -E 'url:\s*http://$PROM_DOMAIN' '$DS_FILE'" >/dev/null 2>&1 || { echo "[ERR] datasource not pointing to $PROM_DOMAIN" >&2; exit 1; }
+echo "[OK] datasource points to domain"
+
+echo "[VERIFY:GRAFANA] bind resolution inside grafana"
+tries=0
+until docker exec argus-grafana getent hosts prom.metric.argus.com >/dev/null 2>&1; do
+  tries=$((tries+1))
+  if (( tries > 24 )); then
+    echo "[ERR] grafana cannot resolve prom.metric.argus.com" >&2
+    exit 1
+  fi
+  echo "[..] waiting DNS propagation in grafana ($tries/24)"; sleep 5
+done
+echo "[OK] domain resolves"
+
+echo "[DONE] grafana verify"
--- a/src/sys/tests/scripts/13_metric_verify_grafana_panels.sh
+++ b/src/sys/tests/scripts/13_metric_verify_grafana_panels.sh
@@ -0,0 +1,87 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+TMP_DIR="$TEST_ROOT/tmp/metric-verify"
+mkdir -p "$TMP_DIR"
+
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+GRAF="http://localhost:${GRAFANA_PORT:-3000}"
+HOSTNAME="${METRIC_TEST_HOSTNAME:-test-metric-node-001}"
+
+echo "[VERIFY:GRAF-PANELS] resolve Prometheus datasource UID via Grafana"
+DS_JSON="$TMP_DIR/graf_ds.json"
+curl -fsS --max-time 10 "$GRAF/api/datasources" >"$DS_JSON"
+DS_UID=$(python3 - "$DS_JSON" <<'PY'
+import json,sys
+arr=json.load(open(sys.argv[1]))
+for ds in arr:
+    if (ds.get('type')=='prometheus'):
+        print(ds.get('uid',''))
+        break
+PY
+)
+if [[ -z "$DS_UID" ]]; then echo "[ERR] no prometheus datasource found in grafana" >&2; exit 1; fi
+echo "[OK] Prometheus DS UID=$DS_UID"
+
+proxy_query() {
+  local q="$1"; local out="$2"
+  curl -fsS --max-time 10 --get "$GRAF/api/datasources/proxy/uid/$DS_UID/api/v1/query" \
+    --data-urlencode "query=$q" >"$out"
+}
+
+assert_vector_recent_nonempty() {
+  local json="$1"; local max_age_sec="${2:-180}"
+  python3 - <<'PY' "$json" "$max_age_sec"
+import json,sys,time
+doc=json.load(open(sys.argv[1]))
+if doc.get('status')!='success':
+    raise SystemExit('prom status != success')
+res=doc.get('data',{}).get('result',[])
+assert res, 'empty result'
+ts=float(res[0]['value'][0])
+assert time.time()-ts < float(sys.argv[2]), f'timestamp too old: {ts}'
+print(int(ts))
+PY
+}
+
+echo "[VERIFY:GRAF-PANELS] Dashboard: Node and GPU Metrics — System Load"
+Q_NODE_LOAD="node_load1{hostname=\"$HOSTNAME\"}"
+proxy_query "$Q_NODE_LOAD" "$TMP_DIR/graf_panel_node_load.json"
+assert_vector_recent_nonempty "$TMP_DIR/graf_panel_node_load.json" 300 >/dev/null
+echo "[OK] node_load1 has recent sample via Grafana proxy"
+
+echo "[VERIFY:GRAF-PANELS] Dashboard: Cluster Dashboard — Node online count"
+Q_NODE_ONLINE='count(count by(hostname) (up{job="node"} == 1))'
+proxy_query "$Q_NODE_ONLINE" "$TMP_DIR/graf_panel_node_online.json"
+python3 - "$TMP_DIR/graf_panel_node_online.json" <<'PY'
+import json,sys
+doc=json.load(open(sys.argv[1]))
+assert doc.get('status')=='success', 'prom status not success'
+res=doc.get('data',{}).get('result',[])
+assert res, 'no series for node online count'
+val=float(res[0]['value'][1])
+assert val>=1, f'node online < 1: {val}'
+print('OK',val)
+PY
+echo "[OK] cluster node online count >= 1 via Grafana proxy"
+
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+# 可选：GPU 面板查询（当启用 GPU 时）
+if [[ "${ENABLE_GPU:-false}" == "true" ]]; then
+  echo "[VERIFY:GRAF-PANELS] GPU Panels — DCGM GPU UTIL"
+  Q_GPU_UTIL='DCGM_FI_DEV_GPU_UTIL'
+  proxy_query "$Q_GPU_UTIL" "$TMP_DIR/graf_panel_dcgm_util.json"
+  assert_vector_recent_nonempty "$TMP_DIR/graf_panel_dcgm_util.json" 300 >/dev/null || { echo "[ERR] dcgm gpu util no recent sample via Grafana proxy" >&2; exit 1; }
+  echo "[OK] dcgm gpu util has recent samples via Grafana proxy"
+fi
+
+echo "[DONE] grafana panels verify"
--- a/src/sys/tests/scripts/13_metric_verify_master.sh
+++ b/src/sys/tests/scripts/13_metric_verify_master.sh
@@ -0,0 +1,110 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+TMP_DIR="$TEST_ROOT/tmp/metric-verify"
+mkdir -p "$TMP_DIR"
+
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+MASTER_BASE="http://localhost:${MASTER_PORT:-32300}/api/v1/master"
+HOSTNAME="${METRIC_TEST_HOSTNAME:-test-metric-node-001}"
+
+curl_json() { curl -fsS --max-time 5 "$1"; }
+
+echo "[VERIFY:MASTER] list nodes and locate target hostname=$HOSTNAME"
+ALL_NODES_JSON="$TMP_DIR/master_nodes.json"
+
+# 重试等待节点出现在 /nodes 列表（最多 120s）
+NODE_ID=""
+for attempt in {1..24}; do
+  curl_json "$MASTER_BASE/nodes" > "$ALL_NODES_JSON" || true
+  NODE_ID=$(python3 - "$ALL_NODES_JSON" "$HOSTNAME" <<'PY'
+import json,sys
+try:
+    nodes=json.load(open(sys.argv[1]))
+except Exception:
+    nodes=[]
+name=sys.argv[2]
+for n in nodes:
+    if n.get('name')==name:
+        print(n.get('id',''))
+        break
+PY
+  )
+  if [[ -n "$NODE_ID" ]]; then break; fi
+  echo "[..] waiting node to appear in /nodes ($attempt/24)"; sleep 5
+done
+
+if [[ -z "$NODE_ID" ]]; then
+  echo "[ERR] master /nodes 中未找到 $HOSTNAME（等待超时）" >&2
+  echo "[HINT] 当前 /nodes 列表如下:" >&2
+  sed -n '1,160p' "$ALL_NODES_JSON" >&2 || true
+  exit 1
+fi
+echo "[OK] node id=$NODE_ID"
+
+echo "[VERIFY:MASTER] get node detail and assert fields"
+DETAIL1_JSON="$TMP_DIR/master_node_${NODE_ID}_detail_1.json"
+curl_json "$MASTER_BASE/nodes/$NODE_ID" > "$DETAIL1_JSON"
+
+# 基础字段与健康项检查（不强制立即 online）
+python3 - "$DETAIL1_JSON" "$HOSTNAME" <<'PY'
+import json,sys,datetime
+j=json.load(open(sys.argv[1]))
+host=sys.argv[2]
+assert j.get('name')==host, f"name mismatch: {j.get('name')} != {host}"
+status=j.get('status')
+assert status in ('initialized','online','offline'), f"unexpected status: {status}"
+md=j.get('meta_data',{})
+assert md.get('hostname',j.get('name'))==host, 'meta_data.hostname mismatch'
+assert 'last_report' in j and j['last_report'], 'last_report missing'
+h=j.get('health',{})
+for key in ('metric-node-exporter','metric-fluent-bit','metric-argus-agent'):
+    if key in h:
+        assert h[key].get('status')=='healthy', f"{key} not healthy: {h[key]}"
+print('OK')
+PY
+
+# 轮询等待 last_report 前进并最终转为 online（最多 90s），容忍短暂 5xx/网络错误
+attempt=0
+T_PRE=0
+until [[ $attempt -ge 18 ]]; do
+  sleep 5
+  DETAIL_CUR="$TMP_DIR/master_node_${NODE_ID}_detail_cur.json"
+  if ! curl_json "$MASTER_BASE/nodes/$NODE_ID" > "$DETAIL_CUR" 2>/dev/null; then
+    echo "[..] retrying node detail fetch ($attempt/18)"; ((attempt++)); continue
+  fi
+  read -r STATUS_CUR T_CUR < <(python3 - "$DETAIL_CUR" <<'PY'
+import json,sys,datetime
+j=json.load(open(sys.argv[1]))
+st=j.get('status','')
+ts=j.get('last_report','')
+if ts.endswith('Z'): ts=ts.replace('Z','+00:00')
+try:
+    t=float(datetime.datetime.fromisoformat(ts).timestamp())
+except Exception:
+    t=0.0
+print(st)
+print(t)
+PY
+  )
+  if awk -v a="$T_PRE" -v b="$T_CUR" 'BEGIN{exit !(b>a)}'; then
+    T_PRE="$T_CUR"
+  fi
+  if [[ "$STATUS_CUR" == "online" ]]; then
+    echo "[OK] status online and last_report progressed"
+    break
+  fi
+  ((attempt++))
+done
+if (( attempt >= 18 )) && [[ "$STATUS_CUR" != "online" ]]; then
+  echo "[WARN] status did not reach online within timeout; continuing"
+fi
+
+echo "$NODE_ID" > "$TMP_DIR/node_id_metric"
+echo "[DONE] master verify"
--- a/src/sys/tests/scripts/13_metric_verify_prometheus.sh
+++ b/src/sys/tests/scripts/13_metric_verify_prometheus.sh
@@ -0,0 +1,198 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+TMP_DIR="$TEST_ROOT/tmp/metric-verify"
+mkdir -p "$TMP_DIR"
+
+# 载入端口变量
+if [[ -f "$TEST_ROOT/.env" ]]; then
+  set -a; source "$TEST_ROOT/.env"; set +a
+fi
+
+PROM_BASE="http://localhost:${PROMETHEUS_PORT:-9090}/api/v1"
+HOSTNAME="${METRIC_TEST_HOSTNAME:-${METRIC_TEST_HOSTNAME_CPU:-test-metric-node-001}}"
+
+nodes_json="$TEST_ROOT/private/argus/metric/prometheus/nodes.json"
+targets_json="$TEST_ROOT/private/argus/metric/prometheus/targets/node_exporter.json"
+
+echo "[VERIFY:PROM] nodes.json present and contains hostname=$HOSTNAME"
+[[ -f "$nodes_json" ]] || { echo "[ERR] $nodes_json missing" >&2; exit 1; }
+python3 - "$nodes_json" "$HOSTNAME" <<'PY'
+import json,sys
+arr=json.load(open(sys.argv[1]))
+host=sys.argv[2]
+assert any((i.get('hostname')==host) for i in arr), f"{host} not found in nodes.json"
+PY
+echo "[OK] nodes.json contains target"
+
+echo "[VERIFY:PROM] file_sd targets exist for nodes.json entries"
+[[ -f "$targets_json" ]] || { echo "[ERR] $targets_json missing" >&2; exit 1; }
+python3 - "$nodes_json" "$targets_json" "$HOSTNAME" >"$TMP_DIR/prom_targets_ip_inst.txt" <<'PY'
+import json,sys
+nodes=json.load(open(sys.argv[1]))
+file_sd=json.load(open(sys.argv[2]))
+host=sys.argv[3]
+targets=set()
+for item in file_sd:
+    for t in item.get('targets',[]): targets.add(t)
+# choose node matching hostname; fallback to first metric user node; otherwise first
+sel = None
+for n in nodes:
+    if n.get('hostname') == host:
+        sel = n
+        break
+if not sel:
+    for n in nodes:
+        if n.get('user_id') == 'metric':
+            sel = n
+            break
+if not sel and nodes:
+    sel = nodes[0]
+if not sel:
+    raise SystemExit('nodes.json empty or no suitable node found')
+ip = sel['ip']
+inst = f"{ip}:9100"
+print(ip)
+print(inst)
+PY
+IP_FIRST=$(sed -n '1p' "$TMP_DIR/prom_targets_ip_inst.txt")
+INSTANCE=$(sed -n '2p' "$TMP_DIR/prom_targets_ip_inst.txt")
+echo "[INFO] expecting instance in file_sd: $INSTANCE"
+
+# 尝试在 Prometheus 容器内主动刷新 targets（可选加速）
+if docker ps --format '{{.Names}}' | grep -q '^argus-prometheus$'; then
+  echo "[..] triggering update_targets inside argus-prometheus"
+  docker exec argus-prometheus bash -lc \
+    'python3 /usr/local/bin/update_targets.py --config /private/argus/metric/prometheus/nodes.json --targets-dir /private/argus/metric/prometheus/targets >/dev/null 2>&1 || true'
+fi
+
+# 给 Prometheus 一次初始 scrape 周期
+sleep 10
+
+# 若短暂未生成，进行重试（最多 180s），期间多次触发刷新
+retry=0
+until jq -r '.[].targets[]' "$targets_json" 2>/dev/null | grep -q "^${IP_FIRST}:9100$"; do
+  if (( retry >= 36 )); then
+    echo "[ERR] ${IP_FIRST}:9100 not present in file_sd after timeout" >&2
+    echo "[HINT] current targets file content:" >&2
+    sed -n '1,200p' "$targets_json" >&2 || true
+    exit 1
+  fi
+  if (( retry % 3 == 0 )) && docker ps --format '{{.Names}}' | grep -q '^argus-prometheus$'; then
+    docker exec argus-prometheus bash -lc \
+      'python3 /usr/local/bin/update_targets.py --config /private/argus/metric/prometheus/nodes.json --targets-dir /private/argus/metric/prometheus/targets >/dev/null 2>&1 || true'
+  fi
+  echo "[..] waiting file_sd refresh ($retry/36)"; sleep 5; ((retry++))
+done
+
+# 改为以 PromQL up 指标作为健康依据，避免 targets 页面状态抖动
+echo "[VERIFY:PROM] up{job=\"node\",ip=\"$IP_FIRST\"} > 0"
+attempt=0
+until (( attempt >= 60 )); do
+  curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=up{job=\"node\",ip=\"$IP_FIRST\"}" > "$TMP_DIR/prom_up_inst_active.json" || true
+  if python3 - "$TMP_DIR/prom_up_inst_active.json" <<'PY'
+import json,sys
+try:
+    j=json.load(open(sys.argv[1]))
+except Exception:
+    raise SystemExit(1)
+res=j.get('data',{}).get('result',[])
+if res:
+    try:
+        val=float(res[0]['value'][1])
+        if val>0: raise SystemExit(0)
+    except Exception:
+        pass
+raise SystemExit(1)
+PY
+  then
+    echo "[OK] up > 0 (control-plane scrape works)"; break
+  fi
+  if (( attempt % 6 == 0 )) && docker ps --format '{{.Names}}' | grep -q '^argus-prometheus$'; then
+    docker exec argus-prometheus bash -lc \
+      'python3 /usr/local/bin/update_targets.py --config /private/argus/metric/prometheus/nodes.json --targets-dir /private/argus/metric/prometheus/targets >/dev/null 2>&1 || true'
+  fi
+  echo "[..] waiting up{job=\"node\",ip=\"$IP_FIRST\"} > 0 ($attempt/60)"; sleep 5; ((attempt++))
+done
+if (( attempt >= 60 )); then
+  echo "[ERR] up{job=\"node\",ip=\"$IP_FIRST\"} did not become > 0" >&2
+  exit 1
+fi
+
+echo "[VERIFY:PROM] instant up query > 0"
+curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=up{job=\"node\",ip=\"$IP_FIRST\"}" > "$TMP_DIR/prom_up_inst.json"
+python3 - "$TMP_DIR/prom_up_inst.json" <<'PY'
+import json,sys
+j=json.load(open(sys.argv[1]))
+res=j.get('data',{}).get('result',[])
+assert res, 'empty result for up{job="node",instance=...}'
+val=float(res[0]['value'][1])
+assert val>0, f"up value not > 0: {val}"
+PY
+echo "[OK] up > 0"
+
+echo "[VERIFY:PROM] count(up{job=\"node\"}==1) >= 1"
+curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=count(up{job=\"node\"}==1)" > "$TMP_DIR/prom_up_count.json"
+python3 - "$TMP_DIR/prom_up_count.json" <<'PY'
+import json,sys
+j=json.load(open(sys.argv[1]))
+res=j.get('data',{}).get('result',[])
+assert res, 'empty result for count(up{job="node"}==1)'
+val=float(res[0]['value'][1])
+assert val>=1, f"count < 1: {val}"
+PY
+echo "[OK] up count satisfied"
+echo "[DONE] prometheus verify"
+
+# ========== GPU 验证（可选） ==========
+if [[ "${ENABLE_GPU:-false}" == "true" ]]; then
+  echo
+  echo "[VERIFY:PROM][GPU] dcgm targets & up metric"
+  GPU_IP_PORT="${METRIC_TEST_DCGM_GPU:-172.31.0.51:9400}"
+  GPU_IP="${GPU_IP_PORT%%:*}"
+
+  # 1) file_sd 目标存在（在 Prometheus 容器内生成的 targets 文件）
+  TARGETS_FILE="$TEST_ROOT/private/argus/metric/prometheus/targets/dcgm_exporter.json"
+  if [[ ! -f "$TARGETS_FILE" ]]; then
+    echo "[ERR] $TARGETS_FILE missing" >&2; exit 1
+  fi
+  if ! jq -r '.[].targets[]' "$TARGETS_FILE" 2>/dev/null | grep -q "^${GPU_IP}:9400$"; then
+    echo "[ERR] dcgm target not found for ${GPU_IP}:9400" >&2
+    exit 1
+  fi
+  echo "[OK] dcgm target present in file_sd"
+
+  # 2) up{job="dcgm", ip=GPU_IP} == 1
+  curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=up{job=\"dcgm\",ip=\"$GPU_IP\"}==1" > "$TMP_DIR/prom_dcgm_up.json"
+  python3 - "$TMP_DIR/prom_dcgm_up.json" <<'PY'
+import json,sys
+j=json.load(open(sys.argv[1]))
+res=j.get('data',{}).get('result',[])
+assert res, 'up==1 empty for dcgm'
+val=float(res[0]['value'][1])
+assert val==1.0, f'up not 1: {val}'
+print('OK')
+PY
+  echo "[OK] up{job=dcgm,ip=$GPU_IP} == 1"
+
+  # 3) 至少一个 GPU 指标存在（优先 DCGM_FI_DEV_GPU_UTIL，若无则尝试 DCGM_FI_DEV_FB_USED）
+  query_one() {
+    local q="$1"; local out="$2"
+    curl -fsS --max-time 5 --get "$PROM_BASE/query" --data-urlencode "query=$q" > "$out"
+    python3 - "$out" <<'PY'
+import json,sys
+j=json.load(open(sys.argv[1]))
+ok=(j.get('status')=='success' and len(j.get('data',{}).get('result',[]))>0)
+raise SystemExit(0 if ok else 1)
+PY
+  }
+  if query_one 'DCGM_FI_DEV_GPU_UTIL' "$TMP_DIR/prom_dcgm_util.json" || query_one 'DCGM_FI_DEV_FB_USED' "$TMP_DIR/prom_dcgm_fb.json"; then
+    echo "[OK] dcgm metrics present"
+  else
+    echo "[ERR] no dcgm metrics found" >&2; exit 1
+  fi
+
+  echo "[DONE] prometheus gpu verify"
+fi
--- a/src/sys/tests/scripts/14_metric_cleanup.sh
+++ b/src/sys/tests/scripts/14_metric_cleanup.sh
@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+TEST_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
+
+FTP_SHARE="$TEST_ROOT/private/argus/metric/ftp/share"
+
+if [[ -d "$FTP_SHARE" ]]; then
+  echo "[SYS-METRIC] 清理 FTP 发布产物..."
+  rm -f "$FTP_SHARE"/argus-metric_*.tar.gz 2>/dev/null || true
+  rm -f "$FTP_SHARE"/LATEST_VERSION 2>/dev/null || true
+  rm -f "$FTP_SHARE"/dns.conf "$FTP_SHARE"/setup.sh 2>/dev/null || true
+else
+  echo "[SYS-METRIC] FTP 目录不存在，跳过清理"
+fi
+
+echo "[SYS-METRIC] Metric 清理完成"
--- a/src/sys/tests/scripts/metric/test-node-entrypoint.sh
+++ b/src/sys/tests/scripts/metric/test-node-entrypoint.sh
@@ -0,0 +1,45 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ARGUS_BUILD_UID=${ARGUS_BUILD_UID:-2133}
+ARGUS_BUILD_GID=${ARGUS_BUILD_GID:-2015}
+AGENT_ROOT=${AGENT_ROOT:-/private/argus/agent}
+PREPARED_FLAG="/tmp/.metric_node_prepared"
+
+export DEBIAN_FRONTEND=${DEBIAN_FRONTEND:-noninteractive}
+
+if [[ ! -f "$PREPARED_FLAG" ]]; then
+  apt-get update -qq
+  apt-get install -y -qq \
+    curl \
+    net-tools \
+    iproute2 \
+    lsof \
+    procps \
+    ca-certificates \
+    gnupg2 || {
+      echo "[metric-node] Failed to install base packages" >&2
+      exit 1
+    }
+
+  mkdir -p "$(dirname "$PREPARED_FLAG")"
+  touch "$PREPARED_FLAG"
+fi
+
+if [[ -n "${TZ:-}" ]]; then
+  ln -snf "/usr/share/zoneinfo/${TZ}" /etc/localtime 2>/dev/null || true
+  echo "$TZ" > /etc/timezone 2>/dev/null || true
+fi
+
+mkdir -p "$AGENT_ROOT"
+chown -R "${ARGUS_BUILD_UID}:${ARGUS_BUILD_GID}" "$AGENT_ROOT" 2>/dev/null || true
+
+if [[ "${METRIC_NODE_ROLE:-cpu}" == "gpu" ]]; then
+  if ! command -v nvidia-smi >/dev/null 2>&1; then
+    echo "[metric-node] nvidia-smi not available but GPU role requested" >&2
+    exit 1
+  fi
+  nvidia-smi || true
+fi
+
+exec "$@"
--- a/src/sys/tests/scripts/node_entrypoint.sh
+++ b/src/sys/tests/scripts/node_entrypoint.sh
@@ -46,7 +46,9 @@ fi
 # Start Fluent Bit in background (will block, so run via bash -lc &)
 if [[ -x /private/start-fluent-bit.sh ]]; then
  log "starting fluent-bit"
-  bash -lc '/private/start-fluent-bit.sh' &
+  sysctl -w fs.inotify.max_user_instances=512 >/dev/null 2>&1 || true
+  sysctl -w fs.inotify.max_user_watches=524288 >/dev/null 2>&1 || true
+  bash -lc 'ulimit -n 65536 || true; exec /private/start-fluent-bit.sh' &
 else
  log "missing /private/start-fluent-bit.sh; fluent-bit will not start"
 fi
@@ -54,4 +56,3 @@ fi
 # Start agent in foreground as runtime user
 log "starting argus-agent"
 exec su -s /bin/bash -c /usr/local/bin/argus-agent "$RUNTIME_USER"
-