PostgreSQL connection pools: budget connections before scaling Node.js
Size PostgreSQL connection pools across Kubernetes replicas, reserve database capacity, measure queueing and prevent connection storms.
Engineering notes
Practical writing about backend systems, databases, messaging, infrastructure, and interface reliability. The recurring theme is simple: measure first, then change the system.
Size PostgreSQL connection pools across Kubernetes replicas, reserve database capacity, measure queueing and prevent connection storms.
Prevent duplicate orders across retries and replicas with PostgreSQL transactions, fingerprints and outbox delivery.
Use TTL jitter, single-flight and safe cross-replica refreshes to stop cache stampedes.
Design startup, readiness and liveness probes for Node.js services without restart storms.
A practical playbook for safe schema changes: compatible releases, backfills, validation, indexes and rollback.
Measure event-loop delay, utilization and CPU profiles before scaling Node.js replicas.
A practical way to evaluate PostgreSQL 18 async I/O with real storage pressure, query evidence, I/O views, concurrency limits, and a rollback plan.
How to use in-place Pod resource resize safely: diagnose pressure, change deliberately, watch cluster headroom, and keep rollback boring.
A production-minded workflow for separating arrival rate, processing cost, partition skew, rebalances, and downstream pressure before scaling.
A production-minded upgrade plan covering runtime changes, native dependencies, load tests, canaries, observability, and rollback.
Why a successfully processed event can still arrive again, and how to design idempotent Kafka consumers around safe side effects.
Design Redis cache keys, expiry, invalidation, and cache-aside behavior around what a reader can safely see.
A practical way to investigate a slow PostgreSQL endpoint with EXPLAIN plans, cardinality, filters, joins, and evidence instead of guessing.
Design Vue interfaces around loading, empty results, failed requests, stale data, and changing user context.