---

name: disk-monitor description: Use when the user asks to check disk status, report on hard disk health, temperature, or SMART data, or when performing routine server maintenance.

Disk Monitor

Overview

This skill provides tools and guidelines for checking the health status of hard drives across the infrastructure using SMART data. It generates human-readable reports and helps identify disks that need replacement or monitoring.

When to Use

• User asks about server disk health, temperature, or SMART data.
• Performing routine server maintenance checks.
• Investigating system performance issues that might be caused by failing disks.

不适用

• 需要修盘、分区、格式化或替换磁盘的具体运维操作
• 没有网络访问 Scrutiny API、也没有本地 SMART 数据来源的环境
• 只想看磁盘容量占用，而不是健康度、温度、寿命或历史趋势

Checking Disk Health

To check the current health of all disks, execute the included script. Because your working directory may vary, you should use the absolute path to the script:

python3 ./scripts/disk-health-check.py

(Run from within the skill directory, or use the absolute path scripts/disk-health-check.py relative to this skill)

Automation & AI Processing

To export structured tabular data containing pure integer measurements (ideal for AI parsing, parsing loops, or monitoring stacks like Zabbix/Prometheus), add the --csv flag:

python3 ./scripts/disk-health-check.py --csv

Understanding the Output

The script outputs a formatted table containing:

• 主机 (Host): The machine hosting the drive.
• 型号 (Model): The specific model of the drive.
• 温度 (Temperature): Current operating temperature. High temperatures (>50°C) will trigger warnings.
• 运行时间 (Power On Hours): Total time the drive has been active.
• 寿命 (Wearout/Percentage Used): For SSDs, this indicates the remaining lifespan.
• 更新时间 (Update Time): The last time SMART data was collected. Ensure this is recent to guarantee accuracy.
• 风险 (Risk):
  • 🔴 Critical: Immediate attention needed (High temp, end of life).
  • 🟡 Warning: Monitor closely (Elevated temp, long run time, low remaining life).
  • 🟢 Normal: Healthy.

Advanced: Scrutiny APIs (Historical Data & Monitoring)

When the user asks for historical data, trend analysis, or raw monitoring data beyond the CLI script, you can directly query the underlying Scrutiny APIs. The default host is typically https://smart.pve.icu.

Endpoint	Method	Purpose & Data Structure
/api/device/{wwn}/details	GET	Historical Data: Best for building temperature/wearout trend charts. Returns data.device (metadata) and data.smart_results (an array of historical SMART snapshots including date, temp, power_on_hours, and attrs).
/api/summary	GET	Global Snapshot: Returns a lightweight dictionary (data.summary) containing the latest status of all monitored disks. Used by the disk-health-check script.
/api/health	GET	Service Health: Returns status of internal components (influxdb, sqlite, frontend). Useful for testing if the monitoring system itself is down.

Example Automation Flow for Trending:

1. Fetch /api/summary to get all WWN keys.
2. Loop through WWNs and fetch /api/device/{wwn}/details.
3. Extract data.smart_results[].date and data.smart_results[].temp (or any attribute from attrs) to build a time-series dataset.

Reporting Guidelines

When reporting disk health to a user, follow these principles:

1. Highlight the Critical: Always lead with the 🔴 Red and 🟡 Yellow alerts. Don't bury them in a wall of text.
2. Context is Key: If a disk has a high temperature but is 100% healthy otherwise, suggest checking the cooling/fans.
3. Verify Data Freshness: Pay close attention to the 更新时间 (Update Time). If a failing disk's data hasn't updated in weeks, explicitly point out that the data is stale and the disk might already be dead.
4. Actionable Advice: Provide clear recommendations (e.g., "Schedule replacement", "Check ventilation").

Checklist

执行前：

• 已确认目标是健康检查，而不是容量统计或磁盘修复
• 已确认数据源可访问（默认 Scrutiny API 或用户指定 URL）
• 已根据场景决定是否需要 --csv 结构化输出

执行后：

• 已优先突出 🔴 / 🟡 风险盘
• 已检查 更新时间 是否足够新
• 已给出可执行建议，而不是只贴原始表格

常见错误

错误做法	正确做法
把健康检查当成磁盘容量检查	仅用此技能处理 SMART / 温度 / 寿命 / 风险状态
忽略 更新时间，直接断言磁盘当前状态	先判断数据是否新鲜，再下结论
只看总通过/失败数量，不看具体风险盘	先列出异常盘，再补总体统计
需要历史趋势时仍只依赖 /api/summary	改查 /api/device/{wwn}/details 获取历史快照