Start Tracing on Day Zero
summary
Tracing uses OpenTelemetry spans with trace ID, span ID, parent ID, start and end times and attributes, and requires wrapping functions to start active spans and call span.addEvent instead of console.log. It binds logs to requests via traceId binds logs to requests, surfaces slow spans for outlier detection, and can feed traces into LLM prompts. The hard parts are widespread refactors and backend tooling (eg spinning up a collector in Docker), so a local dev-first workflow with OpenTelemetry makes production adoption far easier.
AI-written summary · about
chapters (20)
- 0:00in this talk…
- 0:41welcome to tracing on day 0
- 1:35console.log tracing
- 2:10about Laurynas
- 2:40what is tracing? a crash course
- 3:30what does a trace actually look like?
- 4:02spans
- 4:48traces
- 5:17events
- 5:40attributes
- 6:03tracing is helpful!
- 6:39using traceId to bind logs to requests
- 7:31the catch: tracing is hard
- 9:20local dev-first wins
- 9:58clean consoles
- 10:16log any data at will
- 10:27surface errors faster
- 10:49easier on-ramp
- 11:05how can I get into this
- 11:27audience questions and answers
related talks
36mOpenTracing Project Deep Dive
23mIntroduction to OpenTracing
24mObservability and instrumentation via OpenTelemetry
26mOpenTracing Isn't Just Tracing: Measure Twice, Instrument Once
22mDistributed Tracing
22mTracing is for Everyone, Not Just Backend Engineers. (How Tracing Could Help Front-e...
14mCombining Logs and Traces using OpenTelemetry
59mOpenTelemetry will save your day and night