Tencent Cloud APM Full-Link Tracing Implementation and Best Practices

This article introduces Tencent Cloud's Application Performance Monitoring (APM) full-link tracing capabilities and implementation practices. With the development of microservices and container technologies, traditional monitoring methods struggle to meet current application needs. Metrics, link tracing, and logs have become essential for cloud-native applications, requiring a mature modern observability system to understand what happens inside applications.

The article traces the evolution from traditional monitoring to observability, highlighting how OpenTelemetry, born in 2019, unified the Trace model and became the de facto standard in the full-link tracing field. Tencent Cloud APM is built on OpenTelemetry standards, fully compatible with open standards, and supports multiple trace protocols including OpenTelemetry, Jaeger, Skywalking, and Zipkin.

The article discusses challenges encountered when implementing full-link tracing, including protocol compatibility across different frameworks, storage value optimization for massive trace data, and limitations of bytecode enhancement for business code performance issues. It presents technical solutions including protocol conversion gateways, tail sampling mechanisms, and thread profiling to fill calling gaps.

Best practices are demonstrated through three scenarios: from application topology to call chains for error/slow analysis, front-end and back-end integration for full-link interoperability, and combining application logs with call chains for automatic association. The article concludes with an overview of Tencent Cloud Observability Platform (TCOP), which includes multiple sub-products like APM, CAT, RUM, Prometheus & Grafana services, and cloud product basic monitoring.