Review · AISource: Elastic
Elastic’s Linux profiling preview: Setup, requirements and limits
Elastic says its OpenTelemetry Profiling integration collects continuous CPU profiles on Linux through eBPF. It requires no application instrumentation or restarts, according to Elastic. [1] Version 0.0.1 is a technical preview. [1] OpenTelemetry Profiles remains Alpha. [3] Elastic’s October 14, 2026, walkthrough performs the one-time Universal Profiling storage initialization as superuser. It then configures a Fleet policy with samples_per_second=20 and enrolls the Linux agent. [2] To verify collection, Elastic directs users to Infrastructure > Universal Profiling > Flamegraphs. Users select the traffic window and filter process.executable.name : "java"; agent health alone is insufficient. [2] Elastic reports 30,150 PetClinic requests in four minutes at slightly over half a CPU core. [2] The walkthrough does not report a profiling-overhead measurement. [2]
Continuous profiling
Elastic OpenTelemetry Profiling integration 0.0.1
VerdictPromising
The documentation describes CPU profiling for Elastic teams; technical-preview status warrants a controlled trial before production adoption.
- Best for
- Elastic teams with a Basic subscription, Kibana 9.4.0+, Elastic Agent 9.4.0+ with OpenTelemetry support, and amd64 or arm64 Linux hosts running kernel 5.10+ with eBPF enabled [1]
- Pricing
- Basic subscription required; Elastic’s package reference does not specify package pricing or total ingestion and storage costs [1]
Pros
- Collection without application instrumentation or restarts, according to Elastic [1]
- Explicit sampling-rate configuration in Elastic’s walkthrough [2]
- Java process filtering in Elastic’s flamegraph verification steps [2]
- Documented support for amd64 and arm64 Linux hosts [1]
Cons
- Technical-preview package built on an alpha profiling signal [1][3]
- CPU-only coverage without heap or allocation profiles, according to Elastic’s walkthrough [2]
- GraalVM native-image stacks limited to one native frame in Elastic’s walkthrough [2]
- Separate compatible instrumentation required for trace correlation, according to Elastic [2]
- Required host capabilities and permissions not enumerated in Elastic’s package reference [1]
Why it matters
Elastic’s documented workflow gives teams using its observability stack a way to investigate CPU activity. Sparse CPU samples alone cannot explain why an endpoint waited. [1][2] Grafana Pyroscope’s Java agent supports allocation and lock-event profiling. [5] This documentation-based assessment includes no hands-on measurements of profiling overhead or tests of navigation from a slow span to sampled stacks with version 0.0.1. Elastic’s walkthrough says correlation requires separate compatible instrumentation. Its cited Java example uses Stack 8.13 rather than this preview’s required versions. [1][2]
Attribution
Sources
- 1ElasticOpenTelemetry Profiling Integration for Elastic
- 2Elastic Observability LabsContinuous profiling with OpenTelemetry: CPU flamegraphs
- 3OpenTelemetryOpenTelemetry Profiles
- 4OpenTelemetryOpenTelemetry Profiles Enters Public Alpha
- 5Grafana LabsJava profiling documentation
Read the original source for full details and context.
AI
Anthropic moves Claude Opus 5.5 fast mode out of research preview
Anthropic removed access requests and the beta-header requirement for Claude Opus 5.5 fast mode. Default fast mode rate limits cover Start, Build and Scale, and Azure-hosted Microsoft Foundry deployments are supported.
AI
AWS adds four managed Airflow agent skills; upgrade skill supports Provisioned only
AWS added four agent skills for coding assistants to author, debug and test managed Airflow workflows and upgrade environments. Authoring, debugging and testing support Provisioned and Serverless; the upgrade skill supports Provisioned only.
Review · AI
Nemotron Bolt claims a 24GB footprint; its deployment recipes target larger GPUs
NVIDIA positions Nemotron Bolt for local agentic work, including coding, with a stated 24GB memory footprint. Its published GGUF deployment recipes target discrete GPUs with at least 32GB or 48GB of memory, depending on configuration. NVIDIA announced availability on October 15, 2026. The quantized model card lists 74 billion total parameters and 9 billion active parameters. It supports multimodal inputs and text output. It specifies mixed quantization: IQ2-XXS experts, NVFP4 Mamba layers and Q8_0 embeddings and output. The setup instructions use llama.cpp with a separate LiLiCorr drafter for speculative decoding.