Cut log storage costs with two Elasticsearch data tiers instead of four
Elastic benchmarked one day of logs on SSD with everything older moved to frozen searchable snapshots, and it came out 16x cheaper than keeping all of it hot. The ILM policy and the cost model are both in the post.
Temporal Cloud observability in Elastic: 50+ metrics, zero collectors
Elastic scrapes metrics.temporal.io, so the workflow that has been stuck in Running for an hour turns out to be a task queue nobody is polling, and you find that out before you open a single worker log.
Telemetry Policy: change OpenTelemetry sampling and log levels at runtime, no restart
Telemetry Policy says what you want to happen and leaves each component to work out how. Change an OpenTelemetry Java agent's trace sampling to 1% and the JVM keeps serving traffic.
Monitor Supabase in Elastic: dashboards, alert templates, SLO templates, and zero agents
When your Supabase API goes slow, it could be the node, Postgres, the pooler or PostgREST. Elastic tells you which one and shows you the logs from whichever it was.
Native OTLP metrics ingestion on Elastic Cloud Hosted
Send an exponential OpenTelemetry histogram and Elasticsearch keeps the scale and buckets you sent. All four type and temporality combinations work now, and your SDK and Collector config stay exactly as they are.
AI root cause analysis in Elastic Agent Builder that cites its evidence
The new release failed at 27.2%, the old one at 28.2%, so the deploy was never the cause; the agent worked that out in 72 seconds and handed back a trace ID for the failure that was.
Drain Vercel into Elastic: serverless observability with nothing to install
A drain and an API key put Vercel logs, traces, and Speed Insights into Elastic Cloud, where you can follow a slow request from the edge to the Lambda function behind it.
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.
Your AI agent needs an alibi: Observability and audit trails for Agent Builder in Elastic
Elastic 9.5 traces every Agent Builder run as OpenTelemetry spans in your own cluster, so tool calls and token counts are queryable with ES|QL. One workflow step adds the approval record, in a data stream the pipeline cannot rewrite.
AI agent observability for Microsoft Foundry: two env vars, no collector
Set up LLM tracing once and every model call, tool execution and handoff from your Foundry agent arrives in Kibana as one queryable trace, with token counts on each span and code for Agent Framework, LangGraph and Node.js.
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.
Kubernetes observability: SLO templates that turn alerts into error budgets
Two bad rollouts burned 88% of a 30-day error budget while the SLI still read 99.56%. This post adds four SLO templates that bring burn-rate tracking to the OTel-based alert rules from Part 1, no new instrumentation required.