Category: What's New
Articles tagged What's New
LLM tracing in Elastic APM: prompts, responses, and token counts in the span view
In a twenty-call agentic trace, you can see which span is using the most tokens and read the prompt that caused it. Both live in Elastic APM, so there is no second tool to run.
OpenTelemetry Java extensions: customize traces without forking the agent
One JAR, loaded at startup by the OpenTelemetry Java agent, lets you filter health checks, rename spans, add resource attributes, and control sampling with no application code changes.
Elastic Agent now runs as an OpenTelemetry Collector: Less memory overhead, zero config changes
Elastic Agent 9.3 sends logs, metrics and traces through one OTel Collector pipeline, running Beats integrations alongside native OTel sources in a single Fleet-managed agent.
Correlate logs, metrics, and traces in one ES|QL query
Walk through four investigations, from CPU saturation to pod memory pressure, each answered by a single query across signal types.
vLLM Prometheus metrics for self-hosted LLM tuning: TTFT, KV Cache, and GPU Utilization
Tuning a self-hosted vLLM inference using its Prometheus metrics in Elastic Observability — TTFT, KV cache, prefix caching and DCGM GPU counters
Android application monitoring with OpenTelemetry: distributed tracing from tap to backend
EDOT Android adds mobile APM to your Android app with one Gradle dependency: crash reporting, session tracking and distributed tracing visible in Kibana.
From CrashLoopBackOff to OOMKilled with PromQL in Elasticsearch and Kibana
Use PromQL in Elasticsearch and Kibana to move from a CrashLoopBackOff alert to OOMKilled, memory versus the limit, and a verified fix.
CrashLoopBackOff to root cause in seconds: automating the 20-minute Kubernetes investigation with Elastic Observability
Elastic's Kubernetes Experience fires alongside the CrashLoopBackOff alert and delivers a root-cause hypothesis with evidence before you even open it.
You have the IP, you want the hostname: building a lookup processor for OpenTelemetry
Look up any value from YAML, CSV or DNS inside the OpenTelemetry Collector or wire in your own source through a processor Elastic built and shipped to Collector Contrib.
Three clicks from alert to error log: breaking down RED metrics by any span attribute in Elastic Observability
See which pod, deployment or version is driving a RED metrics change by breaking down span attributes in Discover, then trace a failing span to the error log behind it.
How Elastic cut OpenTelemetry tail sampling memory by 65% with disk-backed trace storage
Elastic contributed two features upstream to the OTel Collector's tail sampling processor. The span-ingest strategy lets sampling decisions happen earlier, and Pebble tail storage moves trace buffering to disk. It costs more CPU, but operators can raise decision_wait and num_traces without OOM kills.
