Observability & SRE · 2024 — 2025
No-Code Observability & Automated RCA
An OpenTelemetry, otelcol and OpenSearch pipeline correlating metrics, logs and traces into automated root-cause analysis. Later implemented in the APM module of the product.
- Zero
- Application code changes
- 3 signals
- Metrics · logs · traces
- Shipped
- Into product APM module
The constraint
We needed distributed tracing and observability across the microservice estate with one hard requirement: no application code changes. Asking 75+ services to adopt an SDK was not realistic.
The approach
OpenTelemetry auto-instrumentation produced signal without touching source. The
OpenTelemetry Collector (otelcol) handled routing and processing, and
OpenSearch stored and queried all three signal types.
The work was in the correlation rather than the collection. Metrics show that something degraded, traces show where the request spent its time, and logs show what the service was doing. Joined on trace and span identity, the three produce a first-pass root-cause analysis rather than three dashboards to reconcile by hand.
Where it went
We ran it against our own platform first. It worked well enough that the same approach was implemented in the APM module of TCS Enterprise Manager, so the internal POC became a shipped product capability.