fix(ops): Layer3 health_monitor state dir and Telegram heredocs

Move fail counters under deploy/var/layer3-health, use docker_cmd with
sudo fallback, and refactor alert messages to avoid quote parse errors.

Also fix watch Sparkline rgba gradients, hide empty probe pagination,
and ship pending slot pause/menu UI polish.

deploy/health_monitor.sh

@@ -9,12 +9,44 @@ set -euo pipefail
 
 DEPLOY_DIR="${NEXUS_DEPLOY_DIR:-/opt/nexus}"
 NEXUS_SERVICE="${NEXUS_SERVICE:-nexus}"
-FAIL_FILE="/tmp/nexus_health_fail_count"
-LAST_RESTART_FILE="/tmp/nexus_health_last_restart"
+STATE_DIR="${DEPLOY_DIR}/var/layer3-health"
+FAIL_FILE="${STATE_DIR}/fail_count"
+LAST_RESTART_FILE="${STATE_DIR}/last_restart"
+# Legacy paths (cron user may have created root-owned files under /tmp)
+LEGACY_FAIL_FILE="/tmp/nexus_health_fail_count"
+LEGACY_LAST_RESTART_FILE="/tmp/nexus_health_last_restart"
 MAX_FAIL=3
 RESTART_COOLDOWN_SEC="${NEXUS_HEALTH_RESTART_COOLDOWN:-300}"
 ENV_PROD="${DEPLOY_DIR}/docker/.env.prod"
 
+docker_cmd() {
+    if docker "$@" 2>/dev/null; then
+        return 0
+    fi
+    if command -v sudo >/dev/null 2>&1; then
+        sudo docker "$@" 2>/dev/null
+    else
+        return 1
+    fi
+}
+
+init_layer3_state() {
+    mkdir -p "$STATE_DIR"
+    chmod 755 "$STATE_DIR" 2>/dev/null || true
+    if [[ ! -f "$FAIL_FILE" && -f "$LEGACY_FAIL_FILE" ]]; then
+        cp "$LEGACY_FAIL_FILE" "$FAIL_FILE" 2>/dev/null || true
+    fi
+    if [[ ! -f "$LAST_RESTART_FILE" && -f "$LEGACY_LAST_RESTART_FILE" ]]; then
+        cp "$LEGACY_LAST_RESTART_FILE" "$LAST_RESTART_FILE" 2>/dev/null || true
+    fi
+}
+
+write_state_file() {
+    local path="$1" value="$2"
+    init_layer3_state
+    printf '%s\n' "$value" >"$path" 2>/dev/null || return 1
+}
+
 resolve_publish_port() {
     local port
     port="$(grep -E '^NEXUS_PUBLISH_PORT=' "$ENV_PROD" 2>/dev/null | head -1 | cut -d= -f2- | tr -d '\r" ' || true)"
@@ -25,7 +57,7 @@ PUBLISH_PORT="$(resolve_publish_port)"
 HEALTH_URL="http://127.0.0.1:${PUBLISH_PORT}/health"
 
 resolve_nexus_container() {
-    docker ps --format '{{.Names}}' 2>/dev/null | grep -E 'nexus-prod-nexus|nexus-nexus-1' | head -1 || true
+    docker_cmd ps --format '{{.Names}}' 2>/dev/null | grep -E 'nexus-prod-nexus|nexus-nexus-1' | head -1 || true
 }
 
 read_env_var() {
@@ -50,7 +82,7 @@ load_telegram_from_db() {
     [[ -n "$cname" ]] || return 1
     [[ -n "${BOT_TOKEN:-}" && -n "${CHAT_ID:-}" ]] && return 0
     local out token chat
-    out="$(docker exec "$cname" python3 -c "
+    out="$(docker_cmd exec "$cname" python3 -c "
 import asyncio
 from server.config import settings as s
 from server.infrastructure.database.session import AsyncSessionLocal
@@ -94,7 +126,7 @@ load_telegram_config() {
                 NEXUS_TELEGRAM_BOT_TOKEN=*) BOT_TOKEN="${line#NEXUS_TELEGRAM_BOT_TOKEN=}" ;;
                 NEXUS_TELEGRAM_CHAT_ID=*) CHAT_ID="${line#NEXUS_TELEGRAM_CHAT_ID=}" ;;
             esac
-        done < <(docker exec "$cname" grep -E '^NEXUS_TELEGRAM_' /app/.env 2>/dev/null || true)
+        done < <(docker_cmd exec "$cname" grep -E '^NEXUS_TELEGRAM_' /app/.env 2>/dev/null || true)
         [[ -n "$BOT_TOKEN" && -n "$CHAT_ID" ]] && return 0
         load_telegram_from_db "$cname" || true
     fi
@@ -103,7 +135,7 @@ load_telegram_config() {
 }
 
 send_telegram() {
-    local msg="${1//\\n/$'\n'}" resp http log="${NEXUS_HEALTH_LOG:-/var/log/nexus_health.log}"
+    local msg="${1}" resp http log="${NEXUS_HEALTH_LOG:-/var/log/nexus_health.log}"
     BOT_TOKEN="$(normalize_env_value "${BOT_TOKEN:-}")"
     CHAT_ID="$(normalize_env_value "${CHAT_ID:-}")"
     [[ -n "$BOT_TOKEN" && -n "$CHAT_ID" ]] || return 1
@@ -121,24 +153,25 @@ send_telegram() {
 
 bump_fail_count() {
     local fail=0
+    init_layer3_state
     if command -v flock >/dev/null 2>&1; then
         (
             flock -x 9
             fail="$(cat "$FAIL_FILE" 2>/dev/null || echo 0)"
             fail=$((fail + 1))
-            echo "$fail" > "$FAIL_FILE"
+            write_state_file "$FAIL_FILE" "$fail" || true
         ) 9>"${FAIL_FILE}.lock"
         fail="$(cat "$FAIL_FILE" 2>/dev/null || echo 0)"
     else
         fail="$(cat "$FAIL_FILE" 2>/dev/null || echo 0)"
         fail=$((fail + 1))
-        echo "$fail" > "$FAIL_FILE"
+        write_state_file "$FAIL_FILE" "$fail" || true
     fi
     echo "$fail"
 }
 
 reset_fail_count() {
-    echo "0" > "$FAIL_FILE"
+    write_state_file "$FAIL_FILE" "0" || true
 }
 
 restart_cooldown_active() {
@@ -151,13 +184,13 @@ restart_cooldown_active() {
 }
 
 mark_restart_attempt() {
-    date +%s > "$LAST_RESTART_FILE"
+    write_state_file "$LAST_RESTART_FILE" "$(date +%s)" || true
 }
 
 restart_nexus() {
     local cname="$1" via=""
     if [[ -n "$cname" ]]; then
-        docker restart "$cname" >/dev/null 2>&1 && via="Docker 容器 ${cname}"
+        docker_cmd restart "$cname" >/dev/null 2>&1 && via="Docker 容器 ${cname}"
     elif command -v supervisorctl >/dev/null 2>&1 && supervisorctl status "$NEXUS_SERVICE" >/dev/null 2>&1; then
         supervisorctl restart "$NEXUS_SERVICE" >/dev/null 2>&1 && via="Supervisor ${NEXUS_SERVICE}"
     else
@@ -166,8 +199,84 @@ restart_nexus() {
     echo "$via"
 }
 
+telegram_health_fail_cooldown() {
+    local body
+    body="$(cat <<EOF
+🔴 <b>【Layer 3 宿主机巡检】健康检查连续失败（冷却期内）</b>
+宿主机：${HOST_LABEL}
+检测地址：${HEALTH_URL}
+连续失败：已达 ${MAX_FAIL} 次
+Docker 容器：${NEXUS_CONTAINER:-未识别}
+具体情况：/health 仍不可达，但距上次自动重启不足 ${RESTART_COOLDOWN_SEC} 秒，本次跳过重启以免频繁抖动
+建议操作：查看 docker logs ${NEXUS_CONTAINER:-nexus} 与 ${NEXUS_HEALTH_LOG:-/var/log/nexus_health.log}
+时间：${TS_LABEL}
+EOF
+)"
+    send_telegram "$body" || true
+}
+
+telegram_health_fail_restart() {
+    local body
+    body="$(cat <<EOF
+🔴 <b>【Layer 3 宿主机巡检】Nexus 连续 ${MAX_FAIL} 次无响应</b>
+宿主机：${HOST_LABEL}
+检测地址：${HEALTH_URL}
+连续失败：${FAIL} 次（阈值 ${MAX_FAIL} 次）
+Docker 容器：${NEXUS_CONTAINER:-未识别}
+具体情况：宿主机 cron 无法访问 API 健康端点，平台可能已宕机或端口异常
+正在执行：自动重启 Nexus 服务…
+时间：${TS_LABEL}
+EOF
+)"
+    send_telegram "$body" || true
+}
+
+telegram_health_recovered() {
+    local via="$1" body
+    body="$(cat <<EOF
+🟢 <b>【Layer 3 宿主机巡检】Nexus 已自动恢复</b>
+宿主机：${HOST_LABEL}
+检测地址：${HEALTH_URL}
+恢复方式：${via}
+具体情况：自动重启后 /health 检测通过，服务已恢复在线
+时间：${TS_LABEL}
+EOF
+)"
+    send_telegram "$body" || true
+}
+
+telegram_health_still_down() {
+    local via="$1" body
+    body="$(cat <<EOF
+🔴 <b>【Layer 3 宿主机巡检】自动重启后仍未恢复</b>
+宿主机：${HOST_LABEL}
+检测地址：${HEALTH_URL}
+已尝试：${via}
+具体情况：重启完成后 /health 仍失败，可能存在数据库/Redis 故障或配置错误
+建议操作：立即 SSH 登录排查 docker logs、MySQL、Redis，需人工介入
+时间：${TS_LABEL}
+EOF
+)"
+    send_telegram "$body" || true
+}
+
+telegram_health_restart_failed() {
+    local body
+    body="$(cat <<EOF
+🔴 <b>【Layer 3 宿主机巡检】无法执行自动重启</b>
+宿主机：${HOST_LABEL}
+检测地址：${HEALTH_URL}
+具体情况：未找到可重启的 Docker 容器或 Supervisor 服务（${NEXUS_SERVICE}）
+建议操作：确认 nexus 容器名称与 docker ps 输出，手动执行 docker restart 或 nx update
+时间：${TS_LABEL}
+EOF
+)"
+    send_telegram "$body" || true
+}
+
 NEXUS_CONTAINER="$(resolve_nexus_container)"
 load_telegram_config "$NEXUS_CONTAINER"
+init_layer3_state
 HOST_LABEL="$(hostname -f 2>/dev/null || hostname 2>/dev/null || echo unknown)"
 TS_LABEL="$(date '+%Y-%m-%d %H:%M:%S %Z' 2>/dev/null || date)"
 
@@ -182,21 +291,21 @@ if [[ "$FAIL" -lt "$MAX_FAIL" ]]; then
 fi
 
 if restart_cooldown_active; then
-    send_telegram "🔴 <b>【Layer 3 宿主机巡检】健康检查连续失败（冷却期内）</b>\n宿主机：${HOST_LABEL}\n检测地址：${HEALTH_URL}\n连续失败：已达 ${MAX_FAIL} 次\nDocker 容器：${NEXUS_CONTAINER:-未识别}\n具体情况：/health 仍不可达，但距上次自动重启不足 ${RESTART_COOLDOWN_SEC} 秒，本次跳过重启以免频繁抖动\n建议操作：查看 docker logs ${NEXUS_CONTAINER:-nexus} 与 /var/log/nexus_health.log\n时间：${TS_LABEL}" || true
+    telegram_health_fail_cooldown
     exit 0
 fi
 
-send_telegram "🔴 <b>【Layer 3 宿主机巡检】Nexus 连续 ${MAX_FAIL} 次无响应</b>\n宿主机：${HOST_LABEL}\n检测地址：${HEALTH_URL}\n连续失败：${FAIL} 次（阈值 ${MAX_FAIL} 次）\nDocker 容器：${NEXUS_CONTAINER:-未识别}\n具体情况：宿主机 cron 无法访问 API 健康端点，平台可能已宕机或端口异常\n正在执行：自动重启 Nexus 服务…\n时间：${TS_LABEL}" || true
+telegram_health_fail_restart
 mark_restart_attempt
 
 if VIA="$(restart_nexus "$NEXUS_CONTAINER")"; then
     sleep 5
     if curl -sf --max-time 5 "$HEALTH_URL" > /dev/null 2>&1; then
-        send_telegram "🟢 <b>【Layer 3 宿主机巡检】Nexus 已自动恢复</b>\n宿主机：${HOST_LABEL}\n检测地址：${HEALTH_URL}\n恢复方式：${VIA}\n具体情况：自动重启后 /health 检测通过，服务已恢复在线\n时间：${TS_LABEL}" || true
+        telegram_health_recovered "$VIA"
         reset_fail_count
     else
-        send_telegram "🔴 <b>【Layer 3 宿主机巡检】自动重启后仍未恢复</b>\n宿主机：${HOST_LABEL}\n检测地址：${HEALTH_URL}\n已尝试：${VIA}\n具体情况：重启完成后 /health 仍失败，可能存在数据库/Redis 故障或配置错误\n建议操作：立即 SSH 登录排查 docker logs、MySQL、Redis，需人工介入\n时间：${TS_LABEL}" || true
+        telegram_health_still_down "$VIA"
     fi
 else
-    send_telegram "🔴 <b>【Layer 3 宿主机巡检】无法执行自动重启</b>\n宿主机：${HOST_LABEL}\n检测地址：${HEALTH_URL}\n具体情况：未找到可重启的 Docker 容器或 Supervisor 服务（${NEXUS_SERVICE}）\n建议操作：确认 nexus 容器名称与 docker ps 输出，手动执行 docker restart 或 nx update\n时间：${TS_LABEL}" || true
+    telegram_health_restart_failed
 fi

docs/audit/2026-06-14-layer3-health-monitor-p2-p3-ui.md

@@ -0,0 +1,37 @@
+# 审计 — Layer3 health_monitor P2 + 监测 UI P3
+
+**Changelog**: `docs/changelog/2026-06-14-layer3-health-monitor-p2-p3-ui.md`
+
+## 审计范围
+
+| 文件 | 变更 | 状态 |
+|------|------|------|
+| `deploy/health_monitor.sh` | 状态目录、docker_cmd、heredoc Telegram | ☑ |
+| `tests/test_ops_patrol_sync.py` | bash -n + 路径断言 | ☑ |
+| `frontend/src/utils/echartsColor.ts` | 新建 | ☑ |
+| `WatchSparkline.vue` | rgba 渐变 | ☑ |
+| `WatchProbeRecordsTable.vue` | 空表隐藏分页 | ☑ |
+| `WatchSlotCard.vue` | ⋮ 菜单、暂停态精简（同批部署） | ☑ |
+| `WatchMetricRing.vue` | `dimmed` 占位（同批） | ☑ |
+| `server/background/watch_probe_runner.py` | 删除未使用 import | ☑ |
+
+## Step 3
+
+| H | 规则 | 结论 |
+|---|------|------|
+| H1 | sudo docker | SAFE — 与 `nexus-1panel.sh` 同模式，失败则跳过重启 |
+| H2 | 状态目录权限 | SAFE — 在 DEPLOY_DIR 下，cron 用户可写 |
+| H3 | Telegram HTML | SAFE — heredoc 真实换行，`--data-urlencode` 不变 |
+
+## Closure
+
+| H | 判定 |
+|---|------|
+| H1–H3 | SAFE |
+
+## DoD
+
+- [x] changelog
+- [x] `bash -n` + pytest ops_patrol
+- [x] vite build
+- [ ] 生产 L3 日志 5min 无新错误

docs/audit/2026-06-14-watch-slot-paused-ui-polish.md

@@ -6,7 +6,7 @@
 
 | 文件 | 变更 | 状态 |
 |------|------|------|
-| `WatchSlotCard.vue` | 暂停横幅/圆环占位/恢复按钮；顶栏操作区间距 | ☑ |
+| `WatchSlotCard.vue` | 顶栏 ⋮ 菜单（暂停/恢复/移除）；底栏暂停；暂停态 UI | ☑ |
 | `WatchMetricRing.vue` | `dimmed` 占位态 | ☑ |
 
 ## Step 3 规则扫描
@@ -14,7 +14,7 @@
 | H | 规则 | 结论 |
 |---|------|------|
 | H1 | 无新 API | SAFE — 仍 `emit('monitoring'/'remove')` |
-| H2 | 误触移除 | SAFE — 开关与 ✕ 间距加大 |
+| H2 | 误触移除/暂停 | SAFE — 顶栏无开关；暂停/移除在菜单内；底栏「暂停」为文字链 |
 | H3 | 无障碍 | SAFE — `aria-label`  on 开关与移除 |
 
 ## Closure

docs/changelog/2026-06-14-layer3-health-monitor-p2-p3-ui.md

@@ -0,0 +1,35 @@
+# 2026-06-14 — Layer3 巡检脚本与监测 UI 小修
+
+## 摘要
+
+修复 `health_monitor.sh` 状态文件权限与 Telegram 长串引号问题；Sparkline 渐变兼容 rgb 主题；探针表无数据时隐藏分页。
+
+## 动机
+
+巡检 P2：cron 日志大量 `unexpected EOF` 与 `/tmp` 状态文件 Permission denied，Layer3 自愈可能失效。P3：Sparkline `${color}33` 在非 hex 主题失效；探针表 total=0 仍显示分页。
+
+## 变更
+
+- `deploy/health_monitor.sh`：状态迁至 `${DEPLOY_DIR}/var/layer3-health`；`docker_cmd` 回退 sudo；Telegram 文案改 heredoc
+- `frontend/src/utils/echartsColor.ts` + `WatchSparkline.vue`：`colorWithAlpha`
+- `WatchProbeRecordsTable.vue`：`total>0` 才显示分页
+- `WatchSlotCard.vue` / `WatchMetricRing.vue`：监测槽 UI（菜单、暂停态，同批）
+- `tests/test_ops_patrol_sync.py`：`bash -n` 与状态路径断言
+
+## 涉及文件
+
+见上。
+
+## 迁移 / 重启
+
+部署后 cron 自动使用新脚本；首次运行创建 `var/layer3-health/`。可选清理 `/tmp/nexus_health_*`（脚本会尝试迁移 fail_count）。
+
+## 验证
+
+```bash
+bash -n deploy/health_monitor.sh
+pytest tests/test_ops_patrol_sync.py -q
+cd frontend && npx vite build
+```
+
+生产：观察 5 分钟 `/var/log/nexus_health.log` 无新 EOF/Permission denied。

docs/changelog/2026-06-14-watch-slot-paused-ui-polish.md

@@ -2,7 +2,7 @@
 
 ## 摘要
 
-关闭实时监测后，去掉大号居中开关与虚线框；改为顶栏统一开关、浅色提示条、半透明圆环占位与「恢复实时监测」按钮，与开启态卡片结构一致。
+暂停态仅保留顶栏（服务器名 + 槽位 + ⋮ 菜单）与居中「恢复实时监测」；隐藏时长选择、状态横幅、圆环占位与底栏。
 
 ## 动机
 
@@ -10,7 +10,7 @@
 
 ## 变更
 
-- `WatchSlotCard.vue`：暂停横幅 + 四圆环占位 + 恢复按钮；底栏文案精简
+- `WatchSlotCard.vue`：顶栏 `⋮` 菜单替代开关与 ✕；底栏「暂停」；暂停横幅 + 四圆环占位 + 恢复按钮
 - `WatchMetricRing.vue`：新增 `dimmed` 占位态（灰色细环、不可交互）
 
 ## 涉及文件
@@ -24,4 +24,4 @@
 
 ## 验证
 
-`cd frontend && npx vite build`；开启监测后顶栏开关与 ✕ 间距适中、不易误触。
+`cd frontend && npx vite build`；顶栏仅槽位芯片与 ⋮ 菜单；暂停/移除需点开菜单；底栏可点「暂停」。

frontend/src/components/watch/WatchProbeRecordsTable.vue

@@ -51,6 +51,7 @@ const headers = [
 ]
 
 const pageCount = computed(() => Math.max(1, Math.ceil(total.value / 50)))
+const showPagination = computed(() => total.value > 0)
 
 async function load() {
   loading.value = true
@@ -155,7 +156,7 @@ watch(
       </v-data-table>
     </v-sheet>
 
-    <div class="d-flex justify-center align-center mt-4">
+    <div v-if="showPagination" class="d-flex justify-center align-center mt-4">
       <v-pagination
         v-model="page"
         :length="pageCount"

frontend/src/components/watch/WatchSlotCard.vue

@@ -85,28 +85,44 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
         <v-chip size="x-small" variant="tonal" :color="monitoringOn ? 'info' : 'warning'" label>
           槽 {{ slot.slot_index + 1 }}
         </v-chip>
-        <v-switch
-          :model-value="monitoringOn"
-          density="compact"
-          hide-details
-          color="primary"
-          class="watch-slot-switch"
-          :aria-label="monitoringOn ? '关闭实时监测' : '开启实时监测'"
-          @click.stop
-          @update:model-value="emit('monitoring', slot.slot_index, $event ?? false)"
-        />
+        <v-menu location="bottom end" :close-on-content-click="true">
+          <template #activator="{ props: menuProps }">
             <v-btn
-          icon="mdi-close"
+              v-bind="menuProps"
+              icon="mdi-dots-vertical"
               size="x-small"
               variant="text"
-          class="watch-slot-remove-btn"
-          aria-label="移除此槽"
+              class="watch-slot-menu-btn"
+              aria-label="槽位操作"
+              @click.stop
+            />
+          </template>
+          <v-list density="compact" class="watch-slot-menu-list" min-width="168">
+            <v-list-item
+              v-if="monitoringOn"
+              prepend-icon="mdi-pause-circle-outline"
+              title="暂停实时监测"
+              @click="emit('monitoring', slot.slot_index, false)"
+            />
+            <v-list-item
+              v-else
+              prepend-icon="mdi-play-circle-outline"
+              title="恢复实时监测"
+              @click="emit('monitoring', slot.slot_index, true)"
+            />
+            <v-divider class="my-1" />
+            <v-list-item
+              prepend-icon="mdi-close-circle-outline"
+              title="移除此槽"
+              base-color="error"
               @click="emit('remove', slot.slot_index)"
             />
+          </v-list>
+        </v-menu>
       </div>
     </div>
 
-    <div class="watch-ttl-row mb-2" @click.stop>
+    <div v-if="monitoringOn" class="watch-ttl-row mb-2" @click.stop>
       <span class="text-caption text-medium-emphasis watch-ttl-label">时长</span>
       <div class="watch-ttl-chips d-flex flex-wrap ga-1">
         <v-chip
@@ -129,21 +145,8 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
       class="watch-slot-paused-body flex-grow-1"
       @click.stop
     >
-      <div class="watch-slot-paused-banner d-flex align-center ga-2 mb-3">
-        <v-icon icon="mdi-radar-off" size="18" class="text-medium-emphasis" />
-        <span class="text-body-2 font-weight-medium">实时监测已暂停</span>
-        <v-spacer />
-        <v-chip size="x-small" variant="tonal" color="warning" label>暂停</v-chip>
-      </div>
-
-      <div class="watch-metric-rings watch-metric-rings--paused d-flex justify-space-between mb-3">
-        <WatchMetricRing label="CPU" accent="cpu" :size="56" dimmed />
-        <WatchMetricRing label="内存" accent="mem" :size="56" dimmed />
-        <WatchMetricRing label="硬盘" accent="disk" :size="56" dimmed />
-        <WatchMetricRing label="负载" accent="load" :size="56" dimmed />
-      </div>
-
       <div class="watch-slot-paused-cta text-center">
+        <v-icon icon="mdi-radar-off" size="40" class="text-medium-emphasis mb-3 watch-slot-paused-icon" />
         <v-btn
           size="small"
           variant="tonal"
@@ -153,7 +156,7 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
         >
           恢复实时监测
         </v-btn>
-        <div class="text-caption text-medium-emphasis mt-2 px-2">
+        <div class="text-caption text-medium-emphasis mt-3 px-4">
           槽位与子机绑定保留 · 关闭期间由 Agent 每 60 秒上报
         </div>
       </div>
@@ -268,6 +271,15 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
         剩余 {{ formatRemaining(slot.remaining_sec) }} · {{ watchTtlShortLabel(slotTtlMinutes) }}
       </span>
       <div class="d-flex ga-1">
+        <v-btn
+          v-if="monitoringOn"
+          size="x-small"
+          variant="text"
+          class="watch-slot-pause-btn"
+          @click.stop="emit('monitoring', slot.slot_index, false)"
+        >
+          暂停
+        </v-btn>
         <v-btn size="x-small" variant="text" @click.stop="emit('processes', slot)">进程</v-btn>
         <v-btn size="x-small" variant="text" @click.stop="emit('history', slot)">历史</v-btn>
       </div>
@@ -282,17 +294,6 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
     </div>
     </div>
     </template>
-
-    <div
-      v-if="!monitoringOn"
-      class="d-flex align-center justify-space-between mt-auto pt-2"
-      @click.stop
-    >
-      <span class="text-caption text-medium-emphasis">
-        {{ watchTtlShortLabel(slotTtlMinutes) }} · 槽位保留
-      </span>
-      <v-btn size="x-small" variant="text" @click.stop="emit('history', slot)">历史</v-btn>
-    </div>
   </v-card>
 
   <v-card
@@ -347,6 +348,7 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
 }
 .watch-slot-card--paused {
   cursor: default;
+  min-height: 200px;
 }
 .watch-slot-card--paused:hover {
   border-color: rgba(var(--v-border-color), var(--v-border-opacity)) !important;
@@ -354,37 +356,31 @@ const showSwap = computed(() => (props.slot.metrics?.swap_total_gb ?? 0) > 0)
 .watch-slot-paused-body {
   display: flex;
   flex-direction: column;
+  align-items: center;
+  justify-content: center;
   flex: 1 1 auto;
-  min-height: 0;
-  padding-top: 4px;
-}
-.watch-slot-paused-banner {
-  padding: 6px 10px;
-  border-radius: 10px;
-  background: rgba(var(--v-theme-warning), 0.08);
-  border: 1px solid rgba(var(--v-theme-warning), 0.18);
+  min-height: 140px;
 }
 .watch-slot-paused-cta {
-  margin-top: auto;
-  padding-bottom: 4px;
+  width: 100%;
+  padding: 8px 0 12px;
 }
-.watch-metric-rings--paused {
-  opacity: 0.92;
+.watch-slot-paused-icon {
+  opacity: 0.45;
 }
 .watch-slot-header-actions {
-  gap: 6px;
+  gap: 4px;
   margin-left: 8px;
 }
-.watch-slot-switch {
+.watch-slot-menu-btn {
   flex: 0 0 auto;
-  transform: scale(0.85);
-  margin-right: 4px;
+  opacity: 0.72;
 }
-.watch-slot-switch :deep(.v-selection-control) {
-  min-height: 32px;
+.watch-slot-menu-btn:hover {
+  opacity: 1;
 }
-.watch-slot-remove-btn {
-  margin-left: 2px;
+.watch-slot-pause-btn {
+  color: rgba(var(--v-theme-warning), 0.95);
 }
 .watch-ttl-row {
   display: flex;

frontend/src/components/watch/WatchSparkline.vue

@@ -8,6 +8,7 @@ import { GridComponent, TooltipComponent } from 'echarts/components'
 import VChart from 'vue-echarts'
 import type { WatchSparkPoint } from '@/composables/useWatchPins'
 import { formatDateTimeBeijing } from '@/utils/datetime'
+import { colorWithAlpha } from '@/utils/echartsColor'
 
 use([CanvasRenderer, LineChart, GridComponent, TooltipComponent])
 
@@ -59,8 +60,8 @@ const chartOption = computed(() => {
           x2: 0,
           y2: 1,
           colorStops: [
-            { offset: 0, color: `${color}33` },
-            { offset: 1, color: `${color}05` },
+            { offset: 0, color: colorWithAlpha(color, 0.2) },
+            { offset: 1, color: colorWithAlpha(color, 0.02) },
           ],
         },
       },

frontend/src/utils/echartsColor.ts

@@ -0,0 +1,27 @@
+/** ECharts 渐变用：将 Vuetify 主题色转为带透明度的 rgba */
+export function colorWithAlpha(color: string, alpha: number): string {
+  const a = Math.min(1, Math.max(0, alpha))
+  const raw = color.trim()
+
+  if (raw.startsWith('#')) {
+    let h = raw.slice(1)
+    if (h.length === 3) {
+      h = h.split('').map((c) => c + c).join('')
+    }
+    if (h.length >= 6) {
+      const r = Number.parseInt(h.slice(0, 2), 16)
+      const g = Number.parseInt(h.slice(2, 4), 16)
+      const b = Number.parseInt(h.slice(4, 6), 16)
+      if (!Number.isNaN(r) && !Number.isNaN(g) && !Number.isNaN(b)) {
+        return `rgba(${r},${g},${b},${a})`
+      }
+    }
+  }
+
+  const rgb = raw.match(/^rgba?\(\s*([\d.]+)\s*,\s*([\d.]+)\s*,\s*([\d.]+)/i)
+  if (rgb) {
+    return `rgba(${rgb[1]},${rgb[2]},${rgb[3]},${a})`
+  }
+
+  return raw
+}

server/background/watch_probe_runner.py

@@ -5,7 +5,6 @@ from __future__ import annotations
 import asyncio
 import json
 import logging
-import time
 from collections import defaultdict
 from datetime import datetime, timedelta, timezone
 from typing import Any

tests/test_ops_patrol_sync.py

@@ -1,5 +1,6 @@
 """Tests for Layer-3 ops patrol Telegram env sync."""
 
+import subprocess
 from pathlib import Path
 
 import pytest
@@ -7,6 +8,19 @@ import pytest
 from server.infrastructure.env_file import escape_env_value, read_env_var, upsert_env_vars
 from server.application.services import ops_patrol_sync
 
+REPO_ROOT = Path(__file__).resolve().parents[1]
+HEALTH_MONITOR = REPO_ROOT / "deploy" / "health_monitor.sh"
+
+
+def test_health_monitor_shell_syntax():
+    subprocess.run(["bash", "-n", str(HEALTH_MONITOR)], check=True)
+
+
+def test_health_monitor_uses_deploy_var_state_dir():
+    text = HEALTH_MONITOR.read_text(encoding="utf-8")
+    assert 'STATE_DIR="${DEPLOY_DIR}/var/layer3-health"' in text
+    assert 'FAIL_FILE="${STATE_DIR}/fail_count"' in text
+
 
 def test_escape_env_value_quotes_special_chars():
     assert escape_env_value('abc"def') == '"abc\\"def"'