Agent Skills: Golang Concurrency

Golang Concurrency

Priority: P0 (CRITICAL)

Principles

• Share Memory by Communicating: Use channels or synchronization primitives instead of unprotected shared memory.
• Context Propagation: Always pass ctx to manage cancellation, deadlines, and graceful termination.
• Throttle Database Load: Reserve goroutines for independent external API calls or CPU work. Avoid unconstrained goroutines hitting databases to prevent connection pool exhaustion.
• Prevent Leaks: Never start a goroutine without a deterministic shutdown mechanism.
• Race Detection: Always verify concurrent code with go test -race.

Implementation Workflow

1. Choose primitive — errgroup.Group for parallel tasks with error propagation, channels for pipelines, sync.Mutex for simple shared state.
2. Bound Concurrency with Semaphore — Cap maximum active goroutines using a buffered channel semaphore:

   sem := make(chan struct{}, maxWorkers)
   for _, task := range tasks {
       task := task // prevent loop pointer capture
       sem <- struct{}{}
       g.Go(func() error {
           defer func() { <-sem }()
           return task.Execute(ctx)
       })
   }
   

3. Respect Context Cancellation — Goroutines performing I/O must honor ctx.Done().
4. Avoid Loop Pointer Capture — Capture iteration variables explicitly before launching goroutines.
5. Test with race detector — Run go test -race in local dev and CI.

See ErrGroup and concurrency patterns and context timeout examples

Anti-Patterns

• No DB connection spam: avoid spawning unbounded goroutines to query the database in parallel.
• No goroutine leaks: ensure every goroutine has an exit path via ctx.Done() or channel closure.
• No loop variable pointer trap: never pass the address of a loop variable (&item) into a concurrent closure.
• No unbuffered goroutine spawn storms: always cap concurrency with worker pools or semaphores.
• No bare goroutines: use errgroup or sync.WaitGroup for lifecycle tracking.

References

• Concurrency Patterns
• Context Usage