Merge pull request #303 from IanShaw027/feature/ops-account-health-score

feat(ops): 运维监控功能增强与优化

backend/internal/repository/ops_repo_metrics.go

@@ -296,9 +296,10 @@ INSERT INTO ops_job_heartbeats (
   last_error_at,
   last_error,
   last_duration_ms,
+  last_result,
   updated_at
 ) VALUES (
-  $1,$2,$3,$4,$5,$6,NOW()
+  $1,$2,$3,$4,$5,$6,$7,NOW()
 )
 ON CONFLICT (job_name) DO UPDATE SET
   last_run_at = COALESCE(EXCLUDED.last_run_at, ops_job_heartbeats.last_run_at),
@@ -312,6 +313,10 @@ ON CONFLICT (job_name) DO UPDATE SET
     ELSE COALESCE(EXCLUDED.last_error, ops_job_heartbeats.last_error)
   END,
   last_duration_ms = COALESCE(EXCLUDED.last_duration_ms, ops_job_heartbeats.last_duration_ms),
+  last_result = CASE
+    WHEN EXCLUDED.last_success_at IS NOT NULL THEN COALESCE(EXCLUDED.last_result, ops_job_heartbeats.last_result)
+    ELSE ops_job_heartbeats.last_result
+  END,
   updated_at = NOW()`
 
 	_, err := r.db.ExecContext(
@@ -323,6 +328,7 @@ ON CONFLICT (job_name) DO UPDATE SET
 		opsNullTime(input.LastErrorAt),
 		opsNullString(input.LastError),
 		opsNullInt(input.LastDurationMs),
+		opsNullString(input.LastResult),
 	)
 	return err
 }
@@ -340,6 +346,7 @@ SELECT
   last_error_at,
   last_error,
   last_duration_ms,
+  last_result,
   updated_at
 FROM ops_job_heartbeats
 ORDER BY job_name ASC`
@@ -359,6 +366,8 @@ ORDER BY job_name ASC`
 		var lastError sql.NullString
 		var lastDuration sql.NullInt64
 
+		var lastResult sql.NullString
+
 		if err := rows.Scan(
 			&item.JobName,
 			&lastRun,
@@ -366,6 +375,7 @@ ORDER BY job_name ASC`
 			&lastErrorAt,
 			&lastError,
 			&lastDuration,
+			&lastResult,
 			&item.UpdatedAt,
 		); err != nil {
 			return nil, err
@@ -391,6 +401,10 @@ ORDER BY job_name ASC`
 			v := lastDuration.Int64
 			item.LastDurationMs = &v
 		}
+		if lastResult.Valid {
+			v := lastResult.String
+			item.LastResult = &v
+		}
 
 		out = append(out, &item)
 	}

backend/internal/service/ops_aggregation_service.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"database/sql"
 	"errors"
+	"fmt"
 	"log"
 	"strings"
 	"sync"
@@ -235,11 +236,13 @@ func (s *OpsAggregationService) aggregateHourly() {
 	successAt := finishedAt
 	hbCtx, hbCancel := context.WithTimeout(context.Background(), 2*time.Second)
 	defer hbCancel()
+	result := truncateString(fmt.Sprintf("window=%s..%s", start.Format(time.RFC3339), end.Format(time.RFC3339)), 2048)
 	_ = s.opsRepo.UpsertJobHeartbeat(hbCtx, &OpsUpsertJobHeartbeatInput{
 		JobName:        opsAggHourlyJobName,
 		LastRunAt:      &runAt,
 		LastSuccessAt:  &successAt,
 		LastDurationMs: &dur,
+		LastResult:     &result,
 	})
 }
 
@@ -331,11 +334,13 @@ func (s *OpsAggregationService) aggregateDaily() {
 	successAt := finishedAt
 	hbCtx, hbCancel := context.WithTimeout(context.Background(), 2*time.Second)
 	defer hbCancel()
+	result := truncateString(fmt.Sprintf("window=%s..%s", start.Format(time.RFC3339), end.Format(time.RFC3339)), 2048)
 	_ = s.opsRepo.UpsertJobHeartbeat(hbCtx, &OpsUpsertJobHeartbeatInput{
 		JobName:        opsAggDailyJobName,
 		LastRunAt:      &runAt,
 		LastSuccessAt:  &successAt,
 		LastDurationMs: &dur,
+		LastResult:     &result,
 	})
 }
 

backend/internal/service/ops_alert_evaluator_service.go

@@ -190,6 +190,13 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 		return
 	}
 
+	rulesTotal := len(rules)
+	rulesEnabled := 0
+	rulesEvaluated := 0
+	eventsCreated := 0
+	eventsResolved := 0
+	emailsSent := 0
+
 	now := time.Now().UTC()
 	safeEnd := now.Truncate(time.Minute)
 	if safeEnd.IsZero() {
@@ -205,6 +212,7 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 		if rule == nil || !rule.Enabled || rule.ID <= 0 {
 			continue
 		}
+		rulesEnabled++
 
 		scopePlatform, scopeGroupID, scopeRegion := parseOpsAlertRuleScope(rule.Filters)
 
@@ -220,6 +228,7 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 			s.resetRuleState(rule.ID, now)
 			continue
 		}
+		rulesEvaluated++
 
 		breachedNow := compareMetric(metricValue, rule.Operator, rule.Threshold)
 		required := requiredSustainedBreaches(rule.SustainedMinutes, interval)
@@ -278,8 +287,11 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 				continue
 			}
 
+			eventsCreated++
 			if created != nil && created.ID > 0 {
-				s.maybeSendAlertEmail(ctx, runtimeCfg, rule, created)
+				if s.maybeSendAlertEmail(ctx, runtimeCfg, rule, created) {
+					emailsSent++
+				}
 			}
 			continue
 		}
@@ -289,11 +301,14 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 			resolvedAt := now
 			if err := s.opsRepo.UpdateAlertEventStatus(ctx, activeEvent.ID, OpsAlertStatusResolved, &resolvedAt); err != nil {
 				log.Printf("[OpsAlertEvaluator] resolve event failed (event=%d): %v", activeEvent.ID, err)
+			} else {
+				eventsResolved++
 			}
 		}
 	}
 
-	s.recordHeartbeatSuccess(runAt, time.Since(startedAt))
+	result := truncateString(fmt.Sprintf("rules=%d enabled=%d evaluated=%d created=%d resolved=%d emails_sent=%d", rulesTotal, rulesEnabled, rulesEvaluated, eventsCreated, eventsResolved, emailsSent), 2048)
+	s.recordHeartbeatSuccess(runAt, time.Since(startedAt), result)
 }
 
 func (s *OpsAlertEvaluatorService) pruneRuleStates(rules []*OpsAlertRule) {
@@ -585,32 +600,32 @@ func buildOpsAlertDescription(rule *OpsAlertRule, value float64, windowMinutes i
 	)
 }
 
-func (s *OpsAlertEvaluatorService) maybeSendAlertEmail(ctx context.Context, runtimeCfg *OpsAlertRuntimeSettings, rule *OpsAlertRule, event *OpsAlertEvent) {
+func (s *OpsAlertEvaluatorService) maybeSendAlertEmail(ctx context.Context, runtimeCfg *OpsAlertRuntimeSettings, rule *OpsAlertRule, event *OpsAlertEvent) bool {
 	if s == nil || s.emailService == nil || s.opsService == nil || event == nil || rule == nil {
-		return
+		return false
 	}
 	if event.EmailSent {
-		return
+		return false
 	}
 	if !rule.NotifyEmail {
-		return
+		return false
 	}
 
 	emailCfg, err := s.opsService.GetEmailNotificationConfig(ctx)
 	if err != nil || emailCfg == nil || !emailCfg.Alert.Enabled {
-		return
+		return false
 	}
 
 	if len(emailCfg.Alert.Recipients) == 0 {
-		return
+		return false
 	}
 	if !shouldSendOpsAlertEmailByMinSeverity(strings.TrimSpace(emailCfg.Alert.MinSeverity), strings.TrimSpace(rule.Severity)) {
-		return
+		return false
 	}
 
 	if runtimeCfg != nil && runtimeCfg.Silencing.Enabled {
 		if isOpsAlertSilenced(time.Now().UTC(), rule, event, runtimeCfg.Silencing) {
-			return
+			return false
 		}
 	}
 
@@ -639,6 +654,7 @@ func (s *OpsAlertEvaluatorService) maybeSendAlertEmail(ctx context.Context, runt
 	if anySent {
 		_ = s.opsRepo.UpdateAlertEventEmailSent(context.Background(), event.ID, true)
 	}
+	return anySent
 }
 
 func buildOpsAlertEmailBody(rule *OpsAlertRule, event *OpsAlertEvent) string {
@@ -806,7 +822,7 @@ func (s *OpsAlertEvaluatorService) maybeLogSkip(key string) {
 	log.Printf("[OpsAlertEvaluator] leader lock held by another instance; skipping (key=%q)", key)
 }
 
-func (s *OpsAlertEvaluatorService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration) {
+func (s *OpsAlertEvaluatorService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration, result string) {
 	if s == nil || s.opsRepo == nil {
 		return
 	}
@@ -814,11 +830,17 @@ func (s *OpsAlertEvaluatorService) recordHeartbeatSuccess(runAt time.Time, durat
 	durMs := duration.Milliseconds()
 	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 	defer cancel()
+	msg := strings.TrimSpace(result)
+	if msg == "" {
+		msg = "ok"
+	}
+	msg = truncateString(msg, 2048)
 	_ = s.opsRepo.UpsertJobHeartbeat(ctx, &OpsUpsertJobHeartbeatInput{
 		JobName:        opsAlertEvaluatorJobName,
 		LastRunAt:      &runAt,
 		LastSuccessAt:  &now,
 		LastDurationMs: &durMs,
+		LastResult:     &msg,
 	})
 }
 

backend/internal/service/ops_cleanup_service.go

@@ -149,7 +149,7 @@ func (s *OpsCleanupService) runScheduled() {
 		log.Printf("[OpsCleanup] cleanup failed: %v", err)
 		return
 	}
-	s.recordHeartbeatSuccess(runAt, time.Since(startedAt))
+	s.recordHeartbeatSuccess(runAt, time.Since(startedAt), counts)
 	log.Printf("[OpsCleanup] cleanup complete: %s", counts)
 }
 
@@ -330,12 +330,13 @@ func (s *OpsCleanupService) tryAcquireLeaderLock(ctx context.Context) (func(), b
 	return release, true
 }
 
-func (s *OpsCleanupService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration) {
+func (s *OpsCleanupService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration, counts opsCleanupDeletedCounts) {
 	if s == nil || s.opsRepo == nil {
 		return
 	}
 	now := time.Now().UTC()
 	durMs := duration.Milliseconds()
+	result := truncateString(counts.String(), 2048)
 	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 	defer cancel()
 	_ = s.opsRepo.UpsertJobHeartbeat(ctx, &OpsUpsertJobHeartbeatInput{
@@ -343,6 +344,7 @@ func (s *OpsCleanupService) recordHeartbeatSuccess(runAt time.Time, duration tim
 		LastRunAt:      &runAt,
 		LastSuccessAt:  &now,
 		LastDurationMs: &durMs,
+		LastResult:     &result,
 	})
 }
 

backend/internal/service/ops_port.go

@@ -235,6 +235,9 @@ type OpsUpsertJobHeartbeatInput struct {
 	LastErrorAt    *time.Time
 	LastError      *string
 	LastDurationMs *int64
+
+	// LastResult is an optional human-readable summary of the last successful run.
+	LastResult *string
 }
 
 type OpsJobHeartbeat struct {
@@ -245,6 +248,7 @@ type OpsJobHeartbeat struct {
 	LastErrorAt    *time.Time `json:"last_error_at"`
 	LastError      *string    `json:"last_error"`
 	LastDurationMs *int64     `json:"last_duration_ms"`
+	LastResult     *string    `json:"last_result"`
 
 	UpdatedAt time.Time `json:"updated_at"`
 }

backend/internal/service/ops_scheduled_report_service.go

@@ -177,6 +177,10 @@ func (s *OpsScheduledReportService) runOnce() {
 		return
 	}
 
+	reportsTotal := len(reports)
+	reportsDue := 0
+	sentAttempts := 0
+
 	for _, report := range reports {
 		if report == nil || !report.Enabled {
 			continue
@@ -184,14 +188,18 @@ func (s *OpsScheduledReportService) runOnce() {
 		if report.NextRunAt.After(now) {
 			continue
 		}
+		reportsDue++
 
-		if err := s.runReport(ctx, report, now); err != nil {
+		attempts, err := s.runReport(ctx, report, now)
+		if err != nil {
 			s.recordHeartbeatError(runAt, time.Since(startedAt), err)
 			return
 		}
+		sentAttempts += attempts
 	}
 
-	s.recordHeartbeatSuccess(runAt, time.Since(startedAt))
+	result := truncateString(fmt.Sprintf("reports=%d due=%d send_attempts=%d", reportsTotal, reportsDue, sentAttempts), 2048)
+	s.recordHeartbeatSuccess(runAt, time.Since(startedAt), result)
 }
 
 type opsScheduledReport struct {
@@ -297,9 +305,9 @@ func (s *OpsScheduledReportService) listScheduledReports(ctx context.Context, no
 	return out
 }
 
-func (s *OpsScheduledReportService) runReport(ctx context.Context, report *opsScheduledReport, now time.Time) error {
+func (s *OpsScheduledReportService) runReport(ctx context.Context, report *opsScheduledReport, now time.Time) (int, error) {
 	if s == nil || s.opsService == nil || s.emailService == nil || report == nil {
-		return nil
+		return 0, nil
 	}
 	if ctx == nil {
 		ctx = context.Background()
@@ -310,11 +318,11 @@ func (s *OpsScheduledReportService) runReport(ctx context.Context, report *opsSc
 
 	content, err := s.generateReportHTML(ctx, report, now)
 	if err != nil {
-		return err
+		return 0, err
 	}
 	if strings.TrimSpace(content) == "" {
 		// Skip sending when the report decides not to emit content (e.g., digest below min count).
-		return nil
+		return 0, nil
 	}
 
 	recipients := report.Recipients
@@ -325,22 +333,24 @@ func (s *OpsScheduledReportService) runReport(ctx context.Context, report *opsSc
 		}
 	}
 	if len(recipients) == 0 {
-		return nil
+		return 0, nil
 	}
 
 	subject := fmt.Sprintf("[Ops Report] %s", strings.TrimSpace(report.Name))
 
+	attempts := 0
 	for _, to := range recipients {
 		addr := strings.TrimSpace(to)
 		if addr == "" {
 			continue
 		}
+		attempts++
 		if err := s.emailService.SendEmail(ctx, addr, subject, content); err != nil {
 			// Ignore per-recipient failures; continue best-effort.
 			continue
 		}
 	}
-	return nil
+	return attempts, nil
 }
 
 func (s *OpsScheduledReportService) generateReportHTML(ctx context.Context, report *opsScheduledReport, now time.Time) (string, error) {
@@ -650,7 +660,7 @@ func (s *OpsScheduledReportService) setLastRunAt(ctx context.Context, reportType
 	_ = s.redisClient.Set(ctx, key, strconv.FormatInt(t.UTC().Unix(), 10), 14*24*time.Hour).Err()
 }
 
-func (s *OpsScheduledReportService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration) {
+func (s *OpsScheduledReportService) recordHeartbeatSuccess(runAt time.Time, duration time.Duration, result string) {
 	if s == nil || s.opsService == nil || s.opsService.opsRepo == nil {
 		return
 	}
@@ -658,11 +668,17 @@ func (s *OpsScheduledReportService) recordHeartbeatSuccess(runAt time.Time, dura
 	durMs := duration.Milliseconds()
 	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
 	defer cancel()
+	msg := strings.TrimSpace(result)
+	if msg == "" {
+		msg = "ok"
+	}
+	msg = truncateString(msg, 2048)
 	_ = s.opsService.opsRepo.UpsertJobHeartbeat(ctx, &OpsUpsertJobHeartbeatInput{
 		JobName:        opsScheduledReportJobName,
 		LastRunAt:      &runAt,
 		LastSuccessAt:  &now,
 		LastDurationMs: &durMs,
+		LastResult:     &msg,
 	})
 }
 

backend/migrations/039_ops_job_heartbeats_add_last_result.sql

+-- Add last_result to ops_job_heartbeats for UI job details.
+
+ALTER TABLE IF EXISTS ops_job_heartbeats
+    ADD COLUMN IF NOT EXISTS last_result TEXT;
+
+COMMENT ON COLUMN ops_job_heartbeats.last_result IS 'Last successful run result summary (human readable).';

frontend/src/api/admin/ops.ts

@@ -293,6 +293,7 @@ export interface OpsJobHeartbeat {
   last_error_at?: string | null
   last_error?: string | null
   last_duration_ms?: number | null
+  last_result?: string | null
   updated_at: string
 }
 

frontend/src/i18n/locales/en.ts

@@ -1925,7 +1925,7 @@ export default {
       errors: 'Errors',
       errorRate: 'error_rate:',
       upstreamRate: 'upstream_rate:',
-      latencyDuration: 'Request Duration (ms)',
+      latencyDuration: 'Request Duration',
       ttftLabel: 'TTFT (first_token_ms)',
       p50: 'p50:',
       p90: 'p90:',
@@ -2590,7 +2590,7 @@ export default {
         errors: 'Error statistics, including total errors, error rate, and upstream error rate.',
         upstreamErrors: 'Upstream error statistics, excluding rate limit errors (429/529).',
         latency: 'Request duration statistics, including p50, p90, p95, p99 percentiles.',
-        ttft: 'Time To First Token, measuring the speed of first byte return in streaming responses.',
+        ttft: 'Time To First Token, measuring the speed of first token return in streaming responses.',
         health: 'System health score (0-100), considering SLA, error rate, and resource usage.'
       },
       charts: {

frontend/src/i18n/locales/zh.ts

@@ -2048,6 +2048,7 @@ export default {
       lastRun: '最近运行',
       lastSuccess: '最近成功',
       lastError: '最近错误',
+      result: '结果',
       noData: '暂无数据',
       loadingText: '加载中...',
       ready: '就绪',
@@ -2062,7 +2063,7 @@ export default {
       avgQps: '平均 QPS',
       avgTps: '平均 TPS',
       avgLatency: '平均请求时长',
-      avgTtft: '平均首字延迟',
+      avgTtft: '平均首 Token 延迟',
       exceptions: '异常数',
       requestErrors: '请求错误',
       errorCount: '错误数',
@@ -2073,8 +2074,8 @@ export default {
       errors: '错误',
       errorRate: '错误率：',
       upstreamRate: '上游错误率：',
-      latencyDuration: '请求时长（毫秒）',
-      ttftLabel: '首字延迟（毫秒）',
+      latencyDuration: '请求时长',
+      ttftLabel: '首 Token 延迟（毫秒）',
       p50: 'p50',
       p90: 'p90',
       p95: 'p95',
@@ -2117,7 +2118,12 @@ export default {
         '6h': '近6小时',
         '24h': '近24小时',
         '7d': '近7天',
-        '30d': '近30天'
+        '30d': '近30天',
+        custom: '自定义'
+      },
+      customTimeRange: {
+        startTime: '开始时间',
+        endTime: '结束时间'
       },
       fullscreen: {
         enter: '进入全屏'
@@ -2146,7 +2152,7 @@ export default {
         memoryHigh: '内存使用率偏高 ({usage}%)',
         memoryHighImpact: '内存压力较大，需要关注',
         memoryHighAction: '监控内存趋势，检查是否有内存泄漏',
-        ttftHigh: '首字节时间偏高 ({ttft}ms)',
+        ttftHigh: '首 Token 时间偏高 ({ttft}ms)',
         ttftHighImpact: '用户感知时长增加',
         ttftHighAction: '优化请求处理流程，减少前置逻辑耗时',
         // Error rate diagnostics
@@ -2738,7 +2744,7 @@ export default {
         sla: '服务等级协议达成率，排除业务限制（如余额不足、配额超限）的成功请求占比。',
         errors: '错误统计，包括总错误数、错误率和上游错误率。',
         latency: '请求时长统计，包括 p50、p90、p95、p99 等百分位数。',
-        ttft: '首Token延迟（Time To First Token），衡量流式响应的首字节返回速度。',
+        ttft: '首 Token 延迟（Time To First Token），衡量流式响应的首 Token 返回速度。',
         health: '系统健康评分（0-100），综合考虑 SLA、错误率和资源使用情况。'
       },
       charts: {

frontend/src/views/admin/AccountsView.vue

@@ -414,7 +414,17 @@ const handleScroll = () => {
   menu.show = false
 }
 
-onMounted(async () => { load(); try { const [p, g] = await Promise.all([adminAPI.proxies.getAll(), adminAPI.groups.getAll()]); proxies.value = p; groups.value = g } catch (error) { console.error('Failed to load proxies/groups:', error) }; window.addEventListener('scroll', handleScroll, true) })
+onMounted(async () => {
+  load()
+  try {
+    const [p, g] = await Promise.all([adminAPI.proxies.getAll(), adminAPI.groups.getAll()])
+    proxies.value = p
+    groups.value = g
+  } catch (error) {
+    console.error('Failed to load proxies/groups:', error)
+  }
+  window.addEventListener('scroll', handleScroll, true)
+})
 
 onUnmounted(() => {
   window.removeEventListener('scroll', handleScroll, true)

frontend/src/views/admin/ops/OpsDashboard.vue

@@ -23,10 +23,13 @@
         :auto-refresh-enabled="autoRefreshEnabled"
         :auto-refresh-countdown="autoRefreshCountdown"
         :fullscreen="isFullscreen"
+        :custom-start-time="customStartTime"
+        :custom-end-time="customEndTime"
         @update:time-range="onTimeRangeChange"
         @update:platform="onPlatformChange"
         @update:group="onGroupChange"
         @update:query-mode="onQueryModeChange"
+        @update:custom-time-range="onCustomTimeRangeChange"
         @refresh="fetchData"
         @open-request-details="handleOpenRequestDetails"
         @open-error-details="openErrorDetails"
@@ -39,7 +42,7 @@
       <!-- Row: Concurrency + Throughput -->
       <div v-if="opsEnabled && !(loading && !hasLoadedOnce)" class="grid grid-cols-1 gap-6 lg:grid-cols-3">
         <div class="lg:col-span-1 min-h-[360px]">
-          <OpsConcurrencyCard :platform-filter="platform" :group-id-filter="groupId" />
+          <OpsConcurrencyCard :platform-filter="platform" :group-id-filter="groupId" :refresh-token="dashboardRefreshToken" />
         </div>
         <div class="lg:col-span-2 min-h-[360px]">
           <OpsThroughputTrendChart
@@ -148,8 +151,8 @@ const { t } = useI18n()
 
 const opsEnabled = computed(() => adminSettingsStore.opsMonitoringEnabled)
 
-type TimeRange = '5m' | '30m' | '1h' | '6h' | '24h'
-const allowedTimeRanges = new Set<TimeRange>(['5m', '30m', '1h', '6h', '24h'])
+type TimeRange = '5m' | '30m' | '1h' | '6h' | '24h' | 'custom'
+const allowedTimeRanges = new Set<TimeRange>(['5m', '30m', '1h', '6h', '24h', 'custom'])
 
 type QueryMode = 'auto' | 'raw' | 'preagg'
 const allowedQueryModes = new Set<QueryMode>(['auto', 'raw', 'preagg'])
@@ -163,6 +166,8 @@ const timeRange = ref<TimeRange>('1h')
 const platform = ref<string>('')
 const groupId = ref<number | null>(null)
 const queryMode = ref<QueryMode>('auto')
+const customStartTime = ref<string | null>(null)
+const customEndTime = ref<string | null>(null)
 
 const QUERY_KEYS = {
   timeRange: 'tr',
@@ -347,23 +352,24 @@ const autoRefreshEnabled = ref(false)
 const autoRefreshIntervalMs = ref(30000) // default 30 seconds
 const autoRefreshCountdown = ref(0)
 
-// Auto refresh timer
-const { pause: pauseAutoRefresh, resume: resumeAutoRefresh } = useIntervalFn(
-  () => {
-    if (autoRefreshEnabled.value && opsEnabled.value && !loading.value) {
-      fetchData()
-    }
-  },
-  autoRefreshIntervalMs,
-  { immediate: false }
-)
+// Used to trigger child component refreshes in a single shared cadence.
+const dashboardRefreshToken = ref(0)
 
-// Countdown timer (updates every second)
+// Countdown timer (drives auto refresh; updates every second)
 const { pause: pauseCountdown, resume: resumeCountdown } = useIntervalFn(
   () => {
-    if (autoRefreshEnabled.value && autoRefreshCountdown.value > 0) {
-      autoRefreshCountdown.value--
+    if (!autoRefreshEnabled.value) return
+    if (!opsEnabled.value) return
+    if (loading.value) return
+
+    if (autoRefreshCountdown.value <= 0) {
+      // Fetch immediately when the countdown reaches 0.
+      // fetchData() will reset the countdown to the full interval.
+      fetchData()
+      return
     }
+
+    autoRefreshCountdown.value -= 1
   },
   1000,
   { immediate: false }
@@ -420,6 +426,11 @@ function onTimeRangeChange(v: string | number | boolean | null) {
   timeRange.value = v as TimeRange
 }
 
+function onCustomTimeRangeChange(startTime: string, endTime: string) {
+  customStartTime.value = startTime
+  customEndTime.value = endTime
+}
+
 function onSettingsSaved() {
   loadThresholds()
   fetchData()
@@ -458,18 +469,32 @@ function openError(id: number) {
   showErrorModal.value = true
 }
 
-async function refreshOverviewWithCancel(fetchSeq: number, signal: AbortSignal) {
-  if (!opsEnabled.value) return
-  try {
-    const data = await opsAPI.getDashboardOverview(
-      {
-        time_range: timeRange.value,
+function buildApiParams() {
+  const params: any = {
     platform: platform.value || undefined,
     group_id: groupId.value ?? undefined,
     mode: queryMode.value
-      },
-      { signal }
-    )
+  }
+
+  if (timeRange.value === 'custom') {
+    if (customStartTime.value && customEndTime.value) {
+      params.start_time = customStartTime.value
+      params.end_time = customEndTime.value
+    } else {
+      // Safety fallback: avoid sending time_range=custom (backend may not support it)
+      params.time_range = '1h'
+    }
+  } else {
+    params.time_range = timeRange.value
+  }
+
+  return params
+}
+
+async function refreshOverviewWithCancel(fetchSeq: number, signal: AbortSignal) {
+  if (!opsEnabled.value) return
+  try {
+    const data = await opsAPI.getDashboardOverview(buildApiParams(), { signal })
     if (fetchSeq !== dashboardFetchSeq) return
     overview.value = data
   } catch (err: any) {
@@ -483,15 +508,7 @@ async function refreshThroughputTrendWithCancel(fetchSeq: number, signal: AbortS
   if (!opsEnabled.value) return
   loadingTrend.value = true
   try {
-    const data = await opsAPI.getThroughputTrend(
-      {
-        time_range: timeRange.value,
-        platform: platform.value || undefined,
-        group_id: groupId.value ?? undefined,
-        mode: queryMode.value
-      },
-      { signal }
-    )
+    const data = await opsAPI.getThroughputTrend(buildApiParams(), { signal })
     if (fetchSeq !== dashboardFetchSeq) return
     throughputTrend.value = data
   } catch (err: any) {
@@ -509,15 +526,7 @@ async function refreshLatencyHistogramWithCancel(fetchSeq: number, signal: Abort
   if (!opsEnabled.value) return
   loadingLatency.value = true
   try {
-    const data = await opsAPI.getLatencyHistogram(
-      {
-        time_range: timeRange.value,
-        platform: platform.value || undefined,
-        group_id: groupId.value ?? undefined,
-        mode: queryMode.value
-      },
-      { signal }
-    )
+    const data = await opsAPI.getLatencyHistogram(buildApiParams(), { signal })
     if (fetchSeq !== dashboardFetchSeq) return
     latencyHistogram.value = data
   } catch (err: any) {
@@ -535,15 +544,7 @@ async function refreshErrorTrendWithCancel(fetchSeq: number, signal: AbortSignal
   if (!opsEnabled.value) return
   loadingErrorTrend.value = true
   try {
-    const data = await opsAPI.getErrorTrend(
-      {
-        time_range: timeRange.value,
-        platform: platform.value || undefined,
-        group_id: groupId.value ?? undefined,
-        mode: queryMode.value
-      },
-      { signal }
-    )
+    const data = await opsAPI.getErrorTrend(buildApiParams(), { signal })
     if (fetchSeq !== dashboardFetchSeq) return
     errorTrend.value = data
   } catch (err: any) {
@@ -561,15 +562,7 @@ async function refreshErrorDistributionWithCancel(fetchSeq: number, signal: Abor
   if (!opsEnabled.value) return
   loadingErrorDistribution.value = true
   try {
-    const data = await opsAPI.getErrorDistribution(
-      {
-        time_range: timeRange.value,
-        platform: platform.value || undefined,
-        group_id: groupId.value ?? undefined,
-        mode: queryMode.value
-      },
-      { signal }
-    )
+    const data = await opsAPI.getErrorDistribution(buildApiParams(), { signal })
     if (fetchSeq !== dashboardFetchSeq) return
     errorDistribution.value = data
   } catch (err: any) {
@@ -612,7 +605,12 @@ async function fetchData() {
       refreshErrorDistributionWithCancel(fetchSeq, dashboardFetchController.signal)
     ])
     if (fetchSeq !== dashboardFetchSeq) return
+
     lastUpdated.value = new Date()
+
+    // Trigger child component refreshes using the same cadence as the header.
+    dashboardRefreshToken.value += 1
+
     // Reset auto refresh countdown after successful fetch
     if (autoRefreshEnabled.value) {
       autoRefreshCountdown.value = Math.floor(autoRefreshIntervalMs.value / 1000)
@@ -686,15 +684,14 @@ onMounted(async () => {
 
   // Start auto refresh if enabled
   if (autoRefreshEnabled.value) {
-    resumeAutoRefresh()
     resumeCountdown()
   }
 })
 
 async function loadThresholds() {
   try {
-    const settings = await opsAPI.getAlertRuntimeSettings()
-    metricThresholds.value = settings.thresholds || null
+    const thresholds = await opsAPI.getMetricThresholds()
+    metricThresholds.value = thresholds || null
   } catch (err) {
     console.warn('[OpsDashboard] Failed to load thresholds', err)
     metricThresholds.value = null
@@ -704,7 +701,6 @@ async function loadThresholds() {
 onUnmounted(() => {
   window.removeEventListener('keydown', handleKeydown)
   abortDashboardFetch()
-  pauseAutoRefresh()
   pauseCountdown()
 })
 
@@ -712,10 +708,8 @@ onUnmounted(() => {
 watch(autoRefreshEnabled, (enabled) => {
   if (enabled) {
     autoRefreshCountdown.value = Math.floor(autoRefreshIntervalMs.value / 1000)
-    resumeAutoRefresh()
     resumeCountdown()
   } else {
-    pauseAutoRefresh()
     pauseCountdown()
     autoRefreshCountdown.value = 0
   }

frontend/src/views/admin/ops/components/OpsConcurrencyCard.vue

 <script setup lang="ts">
-import { computed, onMounted, onUnmounted, ref, watch } from 'vue'
+import { computed, ref, watch } from 'vue'
 import { useI18n } from 'vue-i18n'
-import { useIntervalFn } from '@vueuse/core'
 import { opsAPI, type OpsAccountAvailabilityStatsResponse, type OpsConcurrencyStatsResponse } from '@/api/admin/ops'
 
 interface Props {
   platformFilter?: string
   groupIdFilter?: number | null
+  refreshToken: number
 }
 
 const props = withDefaults(defineProps<Props>(), {
@@ -233,15 +233,13 @@ async function loadData() {
   }
 }
 
-// 定期刷新（5秒）
-const { pause: pauseRefresh, resume: resumeRefresh } = useIntervalFn(
+// 刷新节奏由父组件统一控制（OpsDashboard Header 的刷新状态/倒计时）
+watch(
+  () => props.refreshToken,
   () => {
-    if (realtimeEnabled.value) {
+    if (!realtimeEnabled.value) return
     loadData()
   }
-  },
-  5000,
-  { immediate: false }
 )
 
 function getLoadBarClass(loadPct: number): string {
@@ -271,23 +269,15 @@ function formatDuration(seconds: number): string {
   return `${hours}h`
 }
 
-onMounted(() => {
-  loadData()
-  resumeRefresh()
-})
-
-onUnmounted(() => {
-  pauseRefresh()
-})
-
-watch(realtimeEnabled, async (enabled) => {
-  if (!enabled) {
-    pauseRefresh()
-  } else {
-    resumeRefresh()
+watch(
+  () => realtimeEnabled.value,
+  async (enabled) => {
+    if (enabled) {
       await loadData()
     }
-})
+  },
+  { immediate: true }
+)
 </script>
 
 <template>