OpenTelemetry : standard observability 2026

OpenTelemetry (OTel) est le standard CNCF d'observability depuis 2019 (merge OpenTracing + OpenCensus). 2026 : adoption massive entreprises et SaaS. SDKs production-ready, Collector mature, signals traces + metrics + logs unifiés.

TL;DR

- OpenTelemetry : standard CNCF observability agnostic.

- SDKs : 12 langages production-ready.

- Collector : agent/proxy unified telemetry.

- 2026 : 80 % observability vendors compatibles natif.

Pourquoi OTel matters

Avant OTel : chaque vendor (Datadog, New Relic, Dynatrace) avait son agent propriétaire. Lock-in fort. OTel = standard instrumentation neutre → switch vendor sans réinstrumenter.

Composants

SDKs

• Production ready : Java, .NET, Python, JS/TS, Go, Ruby, PHP, Rust, Swift
• Stable : traces, metrics, logs (depuis 2023)
• Auto-instrumentation : frameworks populaires (Express, FastAPI, Spring)

Collector

• Agent ou proxy
• Receive (OTLP, Jaeger, Zipkin, Prometheus)
• Process (batch, filter, sample, transform)
• Export (Datadog, Grafana, vendors)
• Déployable sidecar, daemon, gateway

Protocole OTLP

• HTTP/gRPC standard
• Remplace progressivement Jaeger, Zipkin formats

Semantic Conventions

• Attributes standardisés (http.method, db.system, etc.)
• Permet vendor-portability dashboards

Pattern moderne 2026

`

App (OTel SDK auto-instrumented)

→ OTLP gRPC/HTTP

→ OTel Collector (gateway)

→ Filter, sample, enrich

→ Export :

├── Grafana Cloud (free tier)

├── Datadog (premium)

└── S3 (long-term storage)

`

Instrumentation Next.js

`typescript

// instrumentation.ts (Next.js)

import { NodeSDK } from '@opentelemetry/sdk-node';

import { getNodeAutoInstrumentations } from '@opentelemetry/auto-instrumentations-node';

import { OTLPTraceExporter } from '@opentelemetry/exporter-trace-otlp-http';

const sdk = new NodeSDK({

serviceName: 'kolonell-api',

traceExporter: new OTLPTraceExporter({

url: 'http://otel-collector:4318/v1/traces',

}),

instrumentations: [getNodeAutoInstrumentations()],

});

sdk.start();

`

Adoption 2026

• Vendors compatibles natif : Datadog, New Relic, Honeycomb, Splunk, Dynatrace, Grafana, Lightstep
• CNCF top project : 1500+ contributeurs, 2nd most active après K8s
• Enterprise adoption : 60 %+ entreprises Fortune 500 (CNCF survey 2024)

Use cases Africa

• Migration vendor lock-in : OTel instrumentation neutre
• Multi-cloud : observability cross-AWS/Azure/GCP
• Compliance : control data flow (Collector filter PII)
• Cost optimization : sample dans Collector avant export Datadog

Sampling stratégies

`yaml

# otel-collector-config.yaml

processors:

tail_sampling:

policies:

• name: errors

type: status_code

status_code: { status_codes: [ERROR] }

• name: slow

type: latency

latency: { threshold_ms: 1000 }

• name: probabilistic

type: probabilistic

probabilistic: { sampling_percentage: 1 }

`

FAQ

Q : OTel remplace Prometheus ?

R : Non, complémentaire. OTel SDK envoie metrics OTLP → Prometheus peut être backend. Beaucoup d'orgs gardent Prom + OTel pour traces.

Q : Coût Collector ?

R : Self-host : VPS $20-50/mois pour startup. Hosted (Honeycomb, Lightstep) : $$$ enterprise.

Q : Migration vers OTel ?

R : 1-3 sprints typique. Auto-instrumentation accélère. Replace progressivement vendor SDKs propriétaires.

Conclusion

OpenTelemetry 2026 : standard observability incontesté. Élimine vendor lock-in, permet portability. Pour startups Africa, OTel + Grafana Cloud free tier excellent ratio. Pour enterprise, OTel permet flexibilité multi-vendor + cost optimization via Collector sampling. Future-proof investissement.