🔭LinkedIn Ties Member Clicks to Backend Traces With OpenTelemetry and pageForestId

Your frontend bug might finally have a backend trail

LinkedIn just wired its client analytics to its backend tracing, so one member interaction can be followed from the app all the way down to the data store. Analytics tells you what the member did. Tracing tells you what happened inside services. Until now those lived in different worlds. If you run a frontend on top of a microservice mess, this is the gap you've been patching with grep and prayer. LinkedIn's fix is a pageForestId that sits above individual traces. Each request keeps its own OpenTelemetry traceId, but requests from the same page context get tied together, so you can debug one slow call and also see everything else that shaped the experience. The plumbing is Kafka for transport and ClickHouse for storage, with coordinated head sampling so traces stay consistent across services. LinkedIn estimates up to 90% of some backend tracing volume could be treated as debug-level data, which tells you where the storage bill hides. Client-side spans get batched and flushed when the app backgrounds, and attribute allowlists, denylists, and redaction keep sensitive data out of traces. The real win is control: instrumentation isn't tied to one commercial backend.