Debugging Slow Infrastructure Runs With OpenTelemetry in OpenTofu
summary
OpenTofu adds opt-in OpenTelemetry tracing (enabled by environment variables) and shows exact bottlenecks: a demo with Jaeger/OTLP and a local HTTP backend with an injected 5s delay exposes backend latency, provider download HTTP calls, state serialization overhead and API throttling. Concrete wins include 30 minutes to 8 minutes runtime reduction and memory dropping 52 GB to 4 GB after provider caching; tooling includes Jaeger OTLP, CPU/memory profiling, and JSON logs for automated monitoring.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:04Why slow runs happen
- 2:52Slow deployment causes
- 5:45OpenTofu’s performance approach
- 7:07What is OpenTelemetry and why it matters
- 9:13How OpenTelemetry tracing works
- 10:38Provider slowdowns in traces
- 13:08Live demo: OpenTelemetry traces
- 20:51Profiling and logging tools
- 22:39Q&A
related talks
47mHow I Solved Production Issues with OpenTelemetry (And How You Can Too)
27mWho Ate My Resources? Trace LLM Performance in Real Time With OTel
05mEnabling Product Thinking for Terraform Modules with OpenTelemetry (and OpenTofu)
31mTroubleshooting and Monitoring Production Applications using OpenTelemetry
19mOops my app performance sucks... what if observability could be a game changer?
48mObservability of Your Application
41mModern Application Debugging: An Introduction To OpenTelemetry
39mEfficient Debugging and Logging with OpenTelemetry in Go