Databuff
GitHub
Skip to content

中文  |  English

User Guide · Alerting

Automatically record and fire alerts when metrics go wrong.


Capabilities

CapabilityDescription
Threshold alertsFire when error rate, latency, throughput, etc. cross the line
Change detectionCatch sudden metric shifts
Scheduled evaluationRuns every minute, looking back 5 minutes
Event recordsTrack trigger, recovery, handling (auto-resolved when metrics recover)
AI analysisAsk for root cause directly from alert details

Evaluation mechanics: Architecture · Alerting.


FeaturePath
Detection rulesConfiguration → Alert Config → Detection Rules
Preset rulesDetection Rules → Preset Rules (one-click copy)
Convergence policyConfiguration → Alert Config → Convergence
Silence scheduleConfiguration → Alert Config → Silence
Alert listAlert Center → Alert List

External notifications (Webhook, email, etc.) are not supported yet. See RoadmapStronger alerting → notification integrations.


Workflow

1. Create a Detection Rule

Configuration → Alert Config → Detection Rules → New Rule

Or copy a template from Preset Rules and adjust.

Configurable fields:

  • Scope: service or instance
  • Metrics: error rate, avg latency, P99 latency, request count, etc.
  • Condition: threshold (above/below) or change detection
  • Severity: Info / Warning / Critical
  • Evaluation: follows platform default (every minute)

2. View and Handle Alerts

Alert Center → Alert List

Filter by service, severity, status. Click an alert for details:

  • Abnormal metric trends
  • Related traces and logs
  • (Optional) AI root cause analysis, or Alert Center → Manual Root Cause Analysis for a time-range investigation
  • Handling log

Alerts auto-resolve when metrics recover.

3. Advanced Config

SettingPurpose
ConvergenceMerge similar alerts to reduce list noise
SilenceSuppress alert evaluation during maintenance windows

Working with AI

From alert details or AI Platform:

"order-service error rate alert — help me analyze the cause"

AI queries metrics, traces, and topology automatically. For Agent integration, use MCP tool queryServiceAlarms — see Agent Integration.


FAQ

SymptomAction
No alerts after creating rulesEnsure services have metrics; evaluation runs every minute; verify rule scope (see Docker / K8s ops troubleshooting)
No alerts after Demo installManually enable a rule in Preset Rules first; install demo app for traffic; wait 1–2 evaluation cycles
Too many alertsTune thresholds; add convergence or silence policies