可观测性三大支柱架构

Architecture

可观测性架构图，实现三大支柱（指标、追踪、日志），包括 OpenTelemetry SDK 埋点、OTLP 收集器管道、Prometheus、Jaeger 和 Loki 后端、Grafana 仪表板、SLO/SLI 跟踪和 PagerDuty 告警集成。该模板提供完整的可观测性技术栈参考，展示遥测数据如何从应用埋点流经收集、存储、可视化到事件响应。对于构建生产监控基础设施的 SRE 团队至关重要。

下载 .fz 文件

完整 FlowZap 代码

Application { # Application Services
n1: rectangle label:"Service A"
n2: rectangle label:"Service B"
n3: rectangle label:"Service C"
n4: rectangle label:"OpenTelemetry SDK"
n1.handle(bottom) -> n4.handle(top) [label="Instrument"]
n2.handle(bottom) -> n4.handle(top) [label="Instrument"]
n3.handle(bottom) -> n4.handle(top) [label="Instrument"]
n4.handle(bottom) -> Collector.n5.handle(top) [label="OTLP Export"]
}
Collector { # OpenTelemetry Collector
n5: rectangle label:"Receive Telemetry"
n6: rectangle label:"Process and Batch"
n7: rectangle label:"Export Metrics"
n8: rectangle label:"Export Traces"
n9: rectangle label:"Export Logs"
n5.handle(right) -> n6.handle(left) [label="Pipeline"]
n6.handle(right) -> n7.handle(left) [label="Metrics"]
n6.handle(bottom) -> n8.handle(top) [label="Traces"]
n6.handle(bottom) -> n9.handle(top) [label="Logs"]
n7.handle(bottom) -> Backends.n10.handle(top) [label="Prometheus"]
n8.handle(bottom) -> Backends.n11.handle(top) [label="Jaeger"]
n9.handle(bottom) -> Backends.n12.handle(top) [label="Loki"]
}
Backends { # Observability Backends
n10: rectangle label:"Prometheus (Metrics)"
n11: rectangle label:"Jaeger (Distributed Traces)"
n12: rectangle label:"Loki (Log Aggregation)"
n13: rectangle label:"Alertmanager"
n10.handle(right) -> n13.handle(left) [label="Alert Rules"]
n10.handle(bottom) -> Visualization.n14.handle(top) [label="PromQL"]
n11.handle(bottom) -> Visualization.n14.handle(top) [label="Trace ID"]
n12.handle(bottom) -> Visualization.n14.handle(top) [label="LogQL"]
}
Visualization { # Visualization & Alerting
n14: rectangle label:"Grafana Dashboard"
n15: rectangle label:"SLO/SLI Tracking"
n16: rectangle label:"Incident Response"
n17: rectangle label:"PagerDuty / Slack"
n14.handle(right) -> n15.handle(left) [label="SLO Burn Rate"]
n15.handle(right) -> n16.handle(left) [label="SLO Breach"]
n16.handle(right) -> n17.handle(left) [label="Alert"]
n13.handle(bottom) -> Visualization.n17.handle(top) [label="Fire Alert"]
}

为什么需要这个工作流？

Monitoring alone (dashboards and alerts) tells you something is wrong but not why. The three pillars of observability—metrics, traces, and logs—provide the complete picture: metrics detect anomalies, traces pinpoint which service in a request chain is slow, and logs reveal the specific error details.

工作原理

Step 1: Application services are instrumented with the OpenTelemetry SDK.
Step 2: Telemetry data (metrics, traces, logs) is exported via OTLP to the OpenTelemetry Collector.
Step 3: The collector processes, batches, and routes data to specialized backends.
Step 4: Prometheus stores metrics, Jaeger stores distributed traces, and Loki stores logs.
Step 5: Grafana dashboards provide unified visualization across all three pillars.
Step 6: SLO/SLI tracking and Alertmanager trigger PagerDuty alerts when error budgets are consumed.

替代方案

Vendor-specific solutions (Datadog, New Relic) provide all-in-one observability but at significant cost. ELK stack handles logs but not traces natively. This template shows the open-source observability stack with OpenTelemetry.

Key Facts

Template Name	可观测性三大支柱架构
Category	Architecture
Steps	6 workflow steps
Format	FlowZap Code (.fz file)