#opentelemetry

Heron : Wireshark for AI Agents: passive eBPF observability

Heron là công cụ phân tích mạng thụ động dành cho AI agents, sử dụng eBPF để giải mã lưu lượng TLS (LLM) thành plaintext mà không cần SDK hay proxy. Phiên bản 0.7.0 bổ sung kiến trúc OpenTelemetry-native, lọc tự động sidecar an ninh Claude Code, và xuất sẵn SFT trajectory cho fine-tuning, chạy dưới dạng binary Rust duy nhất.

Lập trình viên phát triển ứng dụng AI nên đọc để khám phá cách Heron giúp theo dõi và debug giao thức TLS bí mật của các agent AI một cách hiệu quả, từ đó tối ưu hóa hiệu suất và bảo mật mà không cần phụ thuộc vào SDK hoặc proxy phức tạp.

Structured Logging Patterns That Actually Survive Production — Daily DevOps & .NET

A production-focused guide to structured logging in .NET covering six key patterns: using LoggerMessage source generators on hot paths for zero-allocation logging, configuring scopes as property bags with explicit provider opt-in, threading a single correlation ID end-to-end via W3C baggage propagation, treating log levels as an operational contract, choosing and tuning the right log sink (Application Insights, Loki, Elasticsearch, Seq), and adopting OpenTelemetry Logs for new services. Includes concrete code examples, a comparison of common sinks, and anti-patterns to avoid such as logging full request bodies and maintaining dual minimum-level configurations.

Don't Wrap OpenTelemetry — You're Probably Hurting More Than Helping

Wrapping the OpenTelemetry API in custom abstractions is a common but harmful pattern. Two specific anti-patterns are examined: forcing heap allocations by requiring collection types in wrapper signatures (bypassing OTel's carefully designed zero-allocation overloads), and lookup wrappers that perform dictionary or mutex-locked lookups on every record call instead of holding instrument references. Beyond performance, wrappers prevent developers from learning OTel itself, create maintenance burden, and complicate debugging. OTel already provides InMemoryExporter for testing, eliminating that common justification. Legitimate wrapper use cases include legacy migration dual-writing and strict governance requirements. The post points to OTel Weaver as an upstream code-generation approach for type-safe instrumentation without runtime wrappers.

Building in the open: How Dynatrace invests in open source to move the industry forward

Dynatrace outlines its open source contributions across several major cloud-native projects. The post covers their deep involvement in OpenTelemetry (governance, technical committee, community management), their role in creating the W3C Trace Context standard for distributed tracing, founding the OpenFeature project for vendor-neutral feature flag management, and donating the Keptn toolkit to the CNCF for observability-driven deployment orchestration. It also introduces OffOn.dev, a vendor-agnostic community space aimed at sustaining open source contributors and attracting new maintainers.

GrafanaCON 2026: On-demand sessions are now available

All GrafanaCON 2026 sessions from Barcelona are now available on demand. Key announcements include Grafana 13 GA (dynamic dashboards, Git Sync, Grafana Advisor), Loki 4.0's Kafka-backed architecture delivering 20x less data scanned and 10x faster queries, k6 v2.0 with AI-assisted testing and agentic workflow integrations, and Pyroscope 2.0 with up to 95% reduction in symbol data. Grafana Assistant expands to OSS and Enterprise with new AI observability capabilities. Community sessions cover real-world deployments at Irish Rail, Google, Drive Terra, ESnet, and scientific research projects including ocean digital twins and nuclear materials research.

ObservabilityCON 2026 is coming to San Francisco!

ObservabilityCON 2026 will be held at Pier 27 in San Francisco from October 19–21, 2026. The event will feature keynotes on Grafana Cloud and AI-powered observability, live demos of agentic tools, hands-on workshops, and networking. Early access registration is available now with limited 50%-off tickets. A call for presentations is open until July 1. For those who can't attend in person, ObservabilityCON on the Road is visiting São Paulo, London, Madrid, and Bengaluru this fall.

Security Observability with RSigma and the LGTM Stack

RSigma v0.9.0 can be wired into a full detection-to-alert pipeline using the Grafana LGTM stack (Loki, Grafana, Mimir) with Alloy as the OpenTelemetry collector. The setup uses Docker Compose with seven components: Helr replays pre-recorded Okta audit events, Alloy fans logs to both RSigma (for real-time Sigma rule evaluation) and Loki (for storage), and RSigma exposes per-rule Prometheus metrics with rule_title and level labels. Grafana alert rules use Go templates to dynamically map Sigma severity levels (critical/high/medium/low) to P1-P4 priorities, enabling automatic alert routing without manual wiring. The Okta cross-tenant impersonation scenario demonstrates the full chain: four ordered detections trigger a temporal_ordered correlation, producing a critical alert that routes to the appropriate notification channel. The entire alerting configuration is version-controlled and provisioned automatically.

Building Jaeger’s ClickHouse backend: 8.6× compression on 10 million spans

Jaeger v2.18.0 introduces ClickHouse as a storage backend for distributed traces. A Jaeger maintainer explains the architectural decisions behind the implementation, including schema design choices for the primary key (sorted by service_name, name, start_time rather than trace_id), typed attribute storage using ClickHouse Nested columns, and materialized views for fast service/operation lookups. Benchmarks on 10 million spans across a single-node deployment show 8.6× compression (reducing ~6 GiB to ~722 MiB), ingestion throughput above 50k spans/sec, trace retrieval averaging ~100ms, and most search queries completing under 50ms. The feature is available in alpha and leverages ClickHouse's columnar OLAP architecture to handle high-throughput append-only writes and fast analytical aggregations without a separate metrics pipeline.

Deprecating OpenCensus compatibility requirements

OpenTelemetry has officially deprecated OpenCensus compatibility requirements in its specification, effective June 2026, following the archival of OpenCensus in July 2023. New SDKs and implementations are no longer required to implement OpenCensus compatibility. Existing shims will continue to receive maintenance for at least one year, with the earliest possible specification removal set for June 2027. Users still relying on OpenCensus shims are advised to begin planning migration to native OpenTelemetry APIs and OTLP-based workflows. Language-specific migration resources are available for Go, Java, JavaScript, and Python.