Databuff
GitHub
Skip to content

中文  |  English

Architecture · Alerting

Read Telemetry Pipeline and Storage first.

Design Intent

Alerting value is not "making noise" — it is detect early, record clearly, and support analysis.


Alerting Pipeline

StageProblem solved
Rule configDefine service scope, metrics, and trigger conditions
Scheduled evaluationPeriodic metric checks to find anomalies
Event recordsCapture trigger, status, and recovery context
AI analysisHelp locate cause using metrics, traces, and topology

Current Implementation

DataBuff currently focuses on alert rules, evaluation, and event records:

StageCapability
RulesFlexible metric selection and condition configuration
EvaluationScheduled automatic runs — no manual trigger
EventsRecord occurrence, status changes, and recovery hints
AnalysisAI can intervene directly after alert fires

Working with AI

Alerts are not the end — they are the starting point for AI troubleshooting:

  • Alerts tell you "something is wrong"
  • AI tells you "what, why, and how big the impact is"

That is the key value of alert + AI synergy — from record to diagnosis in one flow.


Design Principles

PrincipleDescription
Based on real metricsEvaluation reads real data from APM storage
Facts firstEnsure accurate rule evaluation and event records before extras
AI-readyAlert data is directly queryable and analyzable by AI experts