Skip to content

OpenTelemetry

CNCF project providing vendor-neutral APIs, SDKs, and the Collector for generating, collecting, processing, and exporting telemetry. This topic distills the official guidance program — the "Managed Telemetry Platforms for Kubernetes Workloads" blueprint and the three published production reference implementations (Adobe, Mastodon, Skyscanner) — into reusable architecture patterns.

Overview

The OpenTelemetry docs organize their best-practice material in two distinct categories under Guidance:

Category Nature Purpose
Blueprints Living documents, tightly scoped to one environment/problem Solve common OTel adoption challenges; several may need combining
Reference implementations Point-in-time snapshots of real organizations' pipelines Ground blueprints in fact rather than theoretical advice; explicitly not actively maintained

The reference-implementations initiative is driven by the Developer Experience (DevEx) SIG and was announced May 2026 with exactly three published entries. The End-User SIG runs the blueprint program; its "centralized telemetry platform" work item (sig-end-user issue #246) became the published K8s-workloads blueprint via site PR #10107 (merged 2026-06-24). This distinction matters operationally: blueprint guidance evolves with the project, while a reference implementation describes how one organization succeeded at a point in time.

Every claim distilled into this topic survived unanimous 3-vote adversarial verification against the live pages on 2026-08-27. See Architecture for the full pattern breakdown and Operations for deployment recipes.

Reference Implementations

Organization Tagline Signature pattern Scale
Adobe An OTel pipeline designed for simplicity at scale Three-tier pipeline with immutable sidecar + signal-isolated managed namespace Enterprise, voluntary adoption
Mastodon Running Collectors in production with a small team One all-signals Collector per namespace as an Operator CR, deployed by Argo CD ~300k daily actives, ~10M req/min
Skyscanner Managing Collectors across 24 production clusters Single DNS entry point + Istio nearest-available routing over Gateway/Agent tiers 1,000+ microservices, 24 clusters, 6 platform engineers

Evaluation

Quick pattern selection from the three cases plus the blueprint:

  • Platform-engineering org running heterogeneous multi-tenant clusters → follow the managed-platforms blueprint: centralized horizontally scaled Collector Gateways, OTLP-only export, secrets confined to the gateway tier, Operator-first auto-instrumentation.
  • Want zero restart risk for instrumented applications → Adobe's immutable sidecar config is specifically designed so config changes never bounce application pods.
  • Tiny team, global scale, minimal moving parts → Mastodon proves one Collector per namespace behind the Operator + GitOps is "more than sufficient" — no tiers, no custom tooling.
  • Migrating off fragmented vendors to a single backend without lock-in → Skyscanner's vendor-agnostic DNS endpoint keeps collector topology stable across backend swaps (they migrated 300+ services off OpenTracing by bumping one library version).

Position In The Vault

  • OTel Collector — sibling component topic covering the engine internals (pdata, distributions, autoscaling, tail-sampling placement) that these organizational patterns deploy.

The patterns here produce telemetry that every backend topic in this domain consumes:

  • SigNoz — OpenTelemetry-native unified platform on ClickHouse
  • LGTM Stack — Grafana's Tempo/Mimir/Loki receiving OTLP directly
  • Victoria Stack — VictoriaTraces/Logs accepting OTLP as drop-in replacements
  • Apache SkyWalking — accepts OTLP alongside its own agent format
  • Observability 2.0 — the wide-event critique of exactly the three-pillar pipeline this guidance standardizes

Relation to the Domain Hub

The domain hub's Key Concepts section describes the Collector receiver → processor → exporter pipeline generically; this topic covers the deployment topologies (gateway tiers, sidecars, per-namespace CRs) that that pipeline gets packaged into.

Sources

All URLs verified HTTP 200 on 2026-08-27 unless noted.

Source Kind Note
Managed Telemetry Platforms for Kubernetes Workloads Primary — blueprint Core page researched here
Adobe reference implementation Primary — ref impl Published 2026-04-08
Mastodon reference implementation Primary — ref impl Modified 2026-05-02
Skyscanner reference implementation Primary — ref impl Modified 2026-07-14
Guidance index Primary Defines blueprints vs reference implementations
Blueprints index Primary Lists published blueprints
Reference implementations index Primary Confirms exactly three RIs as of 2026-08-27
Introducing OTel Blueprints and Reference Implementations Primary — blog Program announcement, 2026-05-12
DevEx blog: Adobe Primary — blog Companion post
DevEx blog: Mastodon Primary — blog Companion post
DevEx blog: Skyscanner Primary — blog Companion post
Operator automatic instrumentation Primary — docs Verifies annotation names
Skyscanner Engineering on Medium Primary — company post Names New Relic; bot-gated (403 to non-browser clients), open in browser
InfoQ: Skyscanner observability migration Secondary Independent corroboration

Questions

  • Which concrete backends/exporters does Adobe's routing-connector setup support today, and can teams switch backends purely through Helm values? → see Architecture
  • How do these pipelines handle gateway-tier failure modes — persistent queuing, memory-limiter tuning, cross-regional failover when Istio's "nearest available" pool degrades? → see Architecture
  • Have the other two End-User SIG blueprints (non-Kubernetes instrumentation, Kubernetes observability) published since mid-2026, or has a fourth reference implementation joined the list? → re-check the blueprints index