Rolling Out OpenTelemetry

Start small, standardise, expand

Begin with one critical user journey across a few services: deploy a Collector, enable zero-code instrumentation, verify end-to-end traces and context propagation, and connect a backend. Agree on standards early: service naming, resource attributes, sampling policy and attribute conventions. Add manual spans for business-critical paths, connect logs, then expand team by team. Provide shared configuration (a base Collector config, an internal library or environment variable defaults) so every service is consistent.

Wiring a building floor by floor

Electricians agree on plug standards first, then wire one floor, test it, and repeat; they do not rewire the whole building at once.

Measure success by faster investigations

Track time to find the cause of incidents before and after adoption to show value.

Quick check: What is a sensible first step when adopting OpenTelemetry?

• Choose sampling after a year
• Instrument every service manually at once
• Instrument one critical user journey end to end and verify propagation
• Skip the Collector and resource attributes