fix(monitoring): classify host memory pressure accurately

ops/alertmanager/alertmanager.yml

@@ -93,7 +93,7 @@ inhibit_rules:
   - source_matchers:
       - alertname="HostDown"
     target_matchers:
-      - alertname=~"HostHighCpuLoad|HostOutOfMemory|HostOutOfDiskSpace"
+      - alertname=~"HostHighCpuLoad|HostMemoryUsageHigh|HostOutOfMemory|HostOutOfDiskSpace"
     equal: ['host']
 
   - source_matchers:

ops/monitoring/alerts-unified.yml

@@ -104,20 +104,23 @@ groups:
           auto_repair_action: "ssh 192.168.0.{{ $labels.host }} '/home/wooo/scripts/host-sustained-load-controller.py --host {{ $labels.host }} --metrics-file /home/wooo/node_exporter_textfiles/host_runaway_process.prom --docker-stats-file /home/wooo/node_exporter_textfiles/docker_stats.prom --json'"
           runbook: "交給 host-sustained-load-controller 產生 AI controlled packet：orphan browser 走 host-runaway-process-remediation.py dry-run → controlled SIGTERM → verifier；合法 CI/BuildKit 走 runner pressure fail-closed 與 drain/cancel packet；unknown 先跑 host-sustained-load-evidence.py 只讀脫敏證據再選服務專屬 PlayBook；swap 走服務專屬記憶體 PlayBook。禁止直接 docker/systemd/nginx/firewall/reboot。"
 
-      - alert: HostOutOfMemory
+      - alert: HostMemoryUsageHigh
         expr: (1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100 > 85
         for: 5m
         labels:
           severity: warning
-          layer: systemd-188
+          layer: 'systemd-{{ $labels.host }}'
           team: ops
           auto_repair: "true"
           mcp_provider: "ssh_host"
           host_type: "bare_metal"
           alert_category: "host_resource"
         annotations:
-          summary: "主機 {{ $labels.host }} 記憶體不足"
-          description: "記憶體使用率超過 85%"
+          summary: "主機 {{ $labels.host }} 記憶體使用率偏高"
+          description: "MemAvailable 推算使用率超過 85% 持續 5 分鐘；這是容量壓力訊號，不代表已發生 OOM、服務停止或主機關機。"
+          auto_repair_action: >-
+            ssh {{ $labels.instance }} 'free -h; grep "^oom_kill " /proc/vmstat; ps aux --sort=-%mem | head -20'
+          runbook: "唯讀確認 MemAvailable、swap、/proc/vmstat oom_kill、主要記憶體程序與服務健康；禁止 drop_caches、restart、reboot。若 oom_kill 未增加且服務健康，維持 degraded/capacity-pressure，不得誤報 OOM 或主機離線。"
 
       - alert: HostOutOfDiskSpace
         expr: (1 - (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay"} / node_filesystem_size_bytes{fstype!~"tmpfs|overlay"})) * 100 > 85

ops/monitoring/alerts.yml

@@ -72,23 +72,23 @@ groups:
           auto_repair_action: "ssh 192.168.0.{{ $labels.host }} '/home/wooo/scripts/host-sustained-load-controller.py --host {{ $labels.host }} --metrics-file /home/wooo/node_exporter_textfiles/host_runaway_process.prom --docker-stats-file /home/wooo/node_exporter_textfiles/docker_stats.prom --json'"
           runbook: "交給 host-sustained-load-controller 產生 AI controlled packet：orphan browser 走 host-runaway-process-remediation.py dry-run → controlled SIGTERM → verifier；合法 CI/BuildKit 走 runner pressure fail-closed 與 drain/cancel packet；unknown 先跑 host-sustained-load-evidence.py 只讀脫敏證據再選服務專屬 PlayBook；swap 走服務專屬記憶體 PlayBook。禁止直接 docker/systemd/nginx/firewall/reboot。"
 
-      - alert: HostOutOfMemory
+      - alert: HostMemoryUsageHigh
         expr: (1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100 > 85
         for: 5m
         labels:
           severity: warning
-          layer: systemd-188
+          layer: 'systemd-{{ $labels.host }}'
           team: ops
           auto_repair: "true"
           mcp_provider: "ssh_host"
           host_type: "bare_metal"
           alert_category: "host_resource"
         annotations:
-          summary: "主機 {{ $labels.host }} 記憶體不足"
-          description: "記憶體使用率超過 85%"
-          # 2026-05-02 ogt + Claude Sonnet 4.6: 引導 LLM 走 SSH 診斷
-          auto_repair_action: "ssh {{ $labels.instance }} 'ps aux --sort=-%mem | head -20' (host 記憶體診斷；禁 kubectl restart — 主因常為第三方服務)"
-          runbook: "host 記憶體不足排查：SSH 看 top 進程；若為第三方服務需擴容或調 limit"
+          summary: "主機 {{ $labels.host }} 記憶體使用率偏高"
+          description: "MemAvailable 推算使用率超過 85% 持續 5 分鐘；這是容量壓力訊號，不代表已發生 OOM、服務停止或主機關機。"
+          auto_repair_action: >-
+            ssh {{ $labels.instance }} 'free -h; grep "^oom_kill " /proc/vmstat; ps aux --sort=-%mem | head -20'
+          runbook: "唯讀確認 MemAvailable、swap、/proc/vmstat oom_kill、主要記憶體程序與服務健康；禁止 drop_caches、restart、reboot。若 oom_kill 未增加且服務健康，維持 degraded/capacity-pressure，不得誤報 OOM 或主機離線。"
 
       - alert: HostOutOfDiskSpace
         expr: (1 - (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay"} / node_filesystem_size_bytes{fstype!~"tmpfs|overlay"})) * 100 > 85