Observability digest — week of 4 Nov 2026

A big week for OpenTelemetry stability: eBPF instrumentation and the Collector’s transformation language both hit 1.0. Elsewhere, Grafana Labs leaned further into OTLP-native pipelines, and vendors kept pairing observability with AI agents and security.

Releases

• OBI reaches v1.0.0 — OpenTelemetry blog. OpenTelemetry eBPF Instrumentation, the zero-code agent for Linux workloads, is now stable, with documented compatibility promises and new protocols (MQTT, NATS, AMQP, MSSQL and more). Read it for the Config v1 freeze and the migration commands before you upgrade.
• OTTL reaches v1 — OpenTelemetry blog. From Collector Contrib v0.164.0, the transformation language’s syntax and standard functions won’t break across 1.x. New functions now start out in an experimental module. If you have transform processor configs in production, this is the stability guarantee you were waiting for.
• Grafana 13.3 release: all the latest features — Grafana Labs. Cross-dashboard variables, investigation notebooks that pull in panels and Assistant findings, and the first phase of the biggest UI refresh since v7. All are in public preview.
• Bridging Alloy and OpenTelemetry: the OTel engine is now in public preview — Grafana Labs. Alloy 1.21 can run native OTLP Collector pipelines alongside its default engine. The post is useful for deciding which engine fits which workload.
• Trace linkage: connect app traces to database queries — Grafana Labs. Grafana Cloud’s Database Observability now links database spans to query samples and explain plans. It matches on query text, or on exact trace and span IDs if you add sqlcommenter. That is a practical way to close the gap between APM and DB tuning.
• Protobuf checkpoints in the File Log receiver — OpenTelemetry blog. A small but welcome change: checkpoints save and load about five times faster and are roughly 35% smaller, with nothing for you to do.

Deep dives

• LinkedIn uses OpenTelemetry to trace the member journey from tap to backend — InfoQ. LinkedIn links mobile and web actions to backend traces, groups related requests with a page-level ID, and uses coordinated head sampling on a pseudonymised member ID. Kafka carries the telemetry and ClickHouse stores it. This is a solid reference design for client-to-server tracing.
• Keycloak’s observability journey — OpenTelemetry blog. The Keycloak maintainers explain why they moved from the Java agent to native Quarkus integration (startup cost), how they unified metrics and traces, and why push-based telemetry changes the threat model.
• Building your own Collector with only the resource detectors you need — OpenTelemetry blog. A short, hands-on look at using ocb to drop unused resource detectors. The author’s binary shrank about 20% and lost 13 dependency packages, which shrinks your CVE exposure too.

Industry news

• Honeycomb introduces AI SRE agents — PR Newswire. Personal and Team agents, now in early access, work towards standing goals such as watching releases, improving instrumentation and triaging alerts, rather than only reacting to incidents.
• CrowdStrike and Dynatrace join forces on SIEM and observability — Dynatrace. Three integrations: federated search across Falcon and Dynatrace without copying data, shared endpoint and application context for incident response, and Bindplane-based log routing. The security–observability convergence continues.
• Why is Dynatrace stock tumbling today? — Investing.com. Revenue beat and grew 17%, but soft guidance and margin dilution from the Arize acquisition sent shares down about 17% pre-market. It’s a reminder that AI-observability deals aren’t cheap.