What OpenTelemetry Users Taught Us About...
summary
OpenTelemetry users report that getting to production — not hello world — is the hard part, so the SIG collects concrete configs and patterns: Mastodon runs a ~6‑person team, ~300k daily users, and uses one collector per namespace with 0.1% tail sampling plus 100% error traces; Adobe runs thousands of collectors in a three‑tier design (sidecar → deployment collectors per signal → backends) via Helm and the operator; Skyscanner manages 1,000+ services across 24 clusters with a single DNS endpoint, Istio routing, gateway and agent collectors and a base image that disables everything then enables what’s needed. Continuous upgrades and deprecations remain the biggest operational pain.
AI-written summary · about
chapters (12)
- 0:00Introduction and survey
- 1:28Speaker introductions
- 2:11Initial assumptions about DevEx
- 3:58Surveying end users
- 5:45Collecting production stories
- 6:30Mastodon’s OpenTelemetry setup
- 9:22Adobe’s three-tier collector design
- 12:31Skyscanner’s collector architecture
- 14:59Key takeaways from interviews
- 17:06Lessons learned by the SIG
- 18:53OpenTelemetry Blueprints initiative
- 20:17Q&A and closing
