Find articles, news, tools and templates by subject or by the work you need to do.
135 results for “Trace a slow request” · 13–24
News · Platform engineering ·
Google API Gateway’s streaming preview delivers response chunks while a model is still generating. Creation-time settings and request deadlines shape the migration.
Read news: Google API Gateway adds streaming for LLM responsesGuide · Observability ·
A durable queue can preserve telemetry through an outage, then take longer to catch up. Arrival rate, export capacity and event age explain what recovery looks like.
Read guide: OpenTelemetry Collector queues and the wait for fresh dataExplainer · AI in production ·
GANs learn to resemble training data. A cascading-timeout example explains what generated traces contribute to testing and what a controlled fixture supplies.
Read explainer: How GANs generate synthetic telemetryExplainer · Observability ·
Counters, gauges and distributions preserve different information. A misleading latency trend shows why the counted requests matter as much as the chart.
Read explainer: What observability metrics keep and leave outExplainer · AI in production ·
A host that launches codex exec must own the process, thread lease, JSONL trace, cancellation and artifact checks. Resume is a lifecycle decision, not a shell shortcut.
Read explainer: Running Codex CLI as a supervised worker processGuide · AI in production ·
OpenShell can stop an agent after Kubernetes has accepted its scaling request. Recovery then depends on the Deployment state, available replicas and customer-facing result.
Read guide: Recovering an agent task after OpenShell containmentGuide · AI in production ·
A missing response can leave a worker already created. Durable operation IDs and the destination’s idempotency contract let a replacement executor recover the original request.
Read guide: Recovering an automation write after a timeoutNews · Observability ·
CloudWatch Omni combines service discovery, agent evaluations, traces and natural-language queries. Its spaces retain account and Region scope, shaping cross-service investigations.
Read news: CloudWatch Omni brings agent evaluations and telemetry into one workspaceGuide · Platform engineering ·
A disruption budget limits unavailable pods while an inference replica is replaced. GPU placement, model warmup and surviving capacity determine how requests fare during that interval.
Read guide: Kubernetes disruption budgets during GPU maintenanceGuide · AI in production ·
An independent checkout request can contradict an agent’s recovery claim. This evaluation separates the agent’s evidence from the grader’s observations and includes a runnable Python example.
Read guide: Testing an incident-response agent against the service it must recoverNews · Observability ·
Splunk’s September releases combine agent traces, evaluations and token-cost monitoring. A rejected incident summary shows how retries and human corrections affect the apparent saving.
Read news: Splunk connects agent activity, evaluation and cost monitoringExplainer · Platform engineering ·
Power and cooling affect the work a GPU pool can sustain. Device signals, request completions and queue growth help distinguish a physical limit from a workload change.
Read explainer: GPU capacity depends on power and cooling