Engineering Insights

Notes from the trenches. No generic tutorials or high-level summaries. Just hard-won operational understanding, raw metrics, and system realities.

2024 · 12 12 min read

Why Your Service Mesh Is Slower Than You Think

The hidden cost of sidecar proxy interception. Understanding connection pool failures, TLS handshake overheads, and measuring the real latency and CPU tax of Envoy's filter chain under load.

Networking Envoy Istio Latency
2024 · 10 10 min read

The Backpressure Problem No One Talks About

How reactive streams backpressure signals propagate through system boundaries, where they silently fail, and what actually happens to your Kafka lag when a consumer falls behind.

Kafka Streaming Reactive Streams Concurrency
2024 · 08 11 min read

Kubernetes StatefulSet Gotchas in Production

Rolling upgrades on StatefulSets behave very differently from Deployments. Unpacking local PV node locks, PVC lifecycle quirks, and the race conditions that bite you during node drains.

Kubernetes Reliability Storage Operations
2024 · 06 13 min read

Raft Is Not Magic: What the Paper Does Not Explain

Reading the Raft paper gives you the algorithm. Running it in production teaches you about leader lease expiry, partition storms, snapshot transfer interruptions, and follower log divergence.

Consensus Raft Distributed Systems etcd
2024 · 03 11 min read

GC Tuning Is Not About the GC

Most JVM GC problems are allocation problems in disguise. A structured playbook on profiling allocation rates with JFR and async-profiler, and fixing the code before touching G1/ZGC flags.

JVM Performance Garbage Collection Profiling
2024 · 01 10 min read

The Real Cost of Exactly-Once in Kafka

Kafka transactions guarantee exactly-once delivery, but at a price. Unpacking epoch management, producer fencing, read isolation levels, and write latency implications of transactional producers.

Kafka Consistency Transactions Distributed Systems