Channel Semaphore vs x/sync/semaphore: It's Not About Speed
Benchmarking chan struct{} against x/sync/semaphore shows the performance folklore is wrong; the differences that matter are semantic.
Read on MediumBlog
Engineering notes on building reliable systems.
Distributed systems, cloud cost, performance, security platforms, and the engineering choices that make infrastructure easier to operate. Most essays are published on Medium.
Reading guide: Go in Production (11 posts).
errgroup Cancels Less Than You Think
errgroup.WithContext is a tiny API that keeps producing the same bugs. A look at its documented cancellation semantics, which are easy to assume rather than read.
Read on Mediumgoleak Has a 450 Millisecond Attention Span
Why a green goleak test suite can coexist with goroutine counts that climb all week in production, and the narrower question goleak actually answers.
Read on MediumYour Graceful Shutdown Probably Doesn't Drain
srv.Shutdown can return nil while every deploy still spikes 502s. What Go's Shutdown actually promises, and what it leaves for you to handle.
Read on MediumRetries Can Make Outages Worse
Retries are one of the easiest resilience patterns to add and one of the easiest to get wrong, and a badly tuned retry can turn a blip into an outage.
Read on MediumThe Difference Between a Cache and a Lie
A look at why caches are time-limited promises rather than sources of truth, and how explicit invalidation and failure handling keep them honest.
Read on MediumWhy I Prefer Boring Caches
A case for simple, predictable caching strategies that degrade gracefully over clever optimizations that are hard to reason about.
Read on MediumDeterminism Is a Core Contract in Build Caching
A practical look at why deterministic build outputs are essential for reliable caching, and the common sources of nondeterminism to watch for.
Read on MediumRemote Execution, Explained for People Who Just Run make
A practical introduction to remote execution, modeling builds as DAGs of deterministic actions that can be cached and distributed across machines.
Read on MediumGo Slice Initialization Gotchas: The Hole-Creating Bug I Keep Seeing
A look at a common Go bug from mixing make() with a non-zero length and append, and the makezero linter that catches it.
Read on MediumWhy I Always Specify Types for Constants in Go, and Name the Unit Too
A practical note on making Go constants explicit, readable, and harder to misuse by naming both the type and the unit.
Read on MediumContainer-Aware Resource Management in Go: The Problem, Go 1.25's Fix, and What's Still Missing
A practical look at Go resource management inside containers, the Go 1.25 improvements, and the gaps engineers still need to account for.
Read on MediumProfiling a Go Service in Production: pprof in 10 Minutes
A practical guide to enabling pprof, collecting CPU and heap profiles, reading flame graphs, and debugging Go services under real load.
Read on Mediumcontext.Context Is Not Optional: A Practical Guide to Cancellation in Go Services
A practical guide to using context.Context as part of service design, cancellation, deadlines, and shutdown behavior in Go.
Read on MediumGo Interfaces: Why Less Is Almost Always More
A short note on keeping Go interfaces small, behavior-focused, and easier to test and evolve.
Read on MediumThree Go Concurrency Mistakes I See in Almost Every Worker Pool
A practical look at worker-pool mistakes around cancellation, channel ownership, and shutdown paths.
Read on Medium