Explain Non-CPU Latency With Prometheus + proxymock
Diagnose a p95 latency regression that CPU profiling cannot explain using Prometheus histograms, then prove the fix with identical proxymock replays.
Posts about ai/ml
Browse 45 posts in this category
Reproduce A Rare Retry Path With Loki + proxymock
Find the rare payload that sends a service down a retry path, correlate it to the recorded response, and prove the fix with proxymock replay.
Separate Application Latency From Network Failure With Hubble
Use Cilium Hubble flow evidence and proxymock replay to prove whether a Kubernetes timeout came from your code, a dependency, or a dropped packet.
Observe Opaque Services With OpenTelemetry eBPF + proxymock
Discover endpoints, RED metrics, and traces in an opaque Kubernetes service with OpenTelemetry eBPF Instrumentation, then replay traffic with proxymock.
Diagnose Serial N+1 API Calls With Tempo + proxymock
Diagnose serial N+1 API calls with Tempo and proxymock. Replay recorded traffic, inspect trace windows, and prove a concurrency fix preserves behavior.
The Pod Was Cheaper. The Service Wasn’t.
Use OpenCost and proxymock to prove Kubernetes rightsizing lowers cost per successful request without hiding behavior or throughput regressions in testing.
Flamegraphs Find It. Replay Proves It.
Use Grafana Pyroscope and proxymock with an AI coding agent to find a Go CPU hotspot, preserve API behavior, and verify the performance fix.
The Best Model Is a Routing Decision
The best model isn't the smartest — it's whichever should get the next unit of work. Notes from NVIDIA on routing, local models, and the AI factory.
Imaginary Test Data. Real Token Bill.
Testing AI applications with invented traffic looks fine until real users arrive. Then come the retries, the fallback models, and the token bill.