Tech Q&A

Why am I getting 'connection refused' errors in production?

This usually means your connection pool is exhausted. Check your max_connections setting and ensure you're properly releasing connections after use. Also verify that your database server is actually running and accepting connections on the expected port.

How do I increase the connection pool limit?

In your pg_hba.conf, increase max_connections. Also consider using PgBouncer for connection pooling in high-traffic scenarios. PgBouncer sits between your app and PostgreSQL, managing connections more efficiently than your application can.

My container keeps getting OOM killed. How do I debug this?

Run `docker stats` to monitor real-time memory usage. Also check `docker inspect <container_id>` for the memory limit configuration. Look at `/sys/fs/cgroup/memory/memory.max_usage_in_bytes` inside the container for peak usage.

Should I set memory limits on all containers?

Yes, always. Without limits, one runaway container can starve others. Set both memory and memory-swap limits explicitly. A good starting point is 2-4x what the container typically uses.

My API routes timeout after 10 seconds on Vercel. How to fix?

Vercel's Hobby plan has a 10s limit. Upgrade to Pro for 60s, or move long-running tasks to background jobs using Vercel's cron or edge functions. You can also use streaming responses to keep the connection alive.

Can I increase the timeout on self-hosted Next.js?

Yes, set the `maxDuration` export in your route file. For self-hosted, you can also configure your reverse proxy (nginx/Apache) timeout settings. Example: `export const maxDuration = 60;`

How do I migrate a large codebase to strict mode without breaking everything?

Start with `strictNullChecks: false` and enable strict options one at a time. Fix errors file by file using // @ts-expect-error comments temporarily. Consider using the `typescript-strict-plugin` for gradual adoption.

Which strict option should I enable first?

Start with `noImplicitAny`. It catches the most bugs with the least migration effort. Then move to `strictNullChecks`, which is harder but finds more real bugs.

When should I use Redis vs in-memory caching?

Use Redis when you have multiple app instances that need shared cache, or when cache needs to survive restarts. In-memory is fine for single instances with volatile data. Consider latency: in-memory is ~100ns, Redis is ~1ms.

How do I handle cache invalidation?

Use TTL-based expiration for simple cases. For complex invalidation, use cache tags or pub/sub patterns to notify all instances when data changes. The hardest part is knowing WHEN to invalidate—document your invalidation rules clearly.

How do I delete all merged branches at once?

Run: `git branch --merged | grep -v '\*\|main\|master' | xargs -n 1 git branch -d`. This deletes all local branches that have been merged. The grep excludes your current branch and protected branches.

What about remote branches?

First prune: `git fetch --prune`. Then delete specific remote branches: `git push origin --delete branch-name`. For bulk deletion, you'll need a script that lists remote merged branches and deletes them.