Xelvaris

Xelvaris

Practical engineering writing for people who keep systems alive.

Data

Practical Notes on Postgres Connection Pooling

April 19, 2026

Each incoming PostgreSQL connection instantiates an isolated operating system process with dedicated memory overhead, necessitating a dedicated pooling proxy in growing deployments. Classic session pooling acts transparently, whereas high-throughput transaction pooling frequently disrupts stateful features such as advisory locking, pub-sub listeners, and session-scoped variables.

Connection pool instability frequently originates from careless application lifecycle management: unclosed checkouts accumulating over time, open cursors lingering during external API calls, or schema alter scripts monopolizing pooled sockets. Transaction mode exposes these architectural flaws aggressively, failing unhygienic workflows on the spot.

Continue reading →

Operations

Zero-Downtime Deployments Without the Drama

July 11, 2026

Popular engineering lore depicts seamless releases through elaborate blue-green switches and instantaneous traffic flips. Production reality centers on disciplined fundamentals: synthetic probes verifying realistic workloads, graceful connection draining on retiring nodes, and database schema mutati…

Networking

Structuring DNS for Reliability

August 9, 2026

Nameserver outages are devastating because they paralyze systems globally: infrastructure monitors report operational green across every host while end users cannot locate your endpoints. Mitigating this catastrophic blind spot requires standard multi-provider redundancy backed by distinct network b…

Engineering

The Hidden Cost of Chatty Microservices

July 5, 2026

Decomposing monolithic applications substitutes in-memory execution paths with network boundaries, creating massive fan-out overhead. Where an operation once invoked three internal functions, it now executes three network dispatches across distinct services, multiplying timeout policies and failure …

Operations

What Good Observability Actually Looks Like

May 4, 2026

Dashboards multiply like rabbits and answer questions nobody asked this month. Good observability starts from the opposite direction: the alert fired, a human is awake, and the tooling should answer 'what changed' within a minute.…

More reading

About us

We cover the unglamorous middle of software: queues that back up, caches that lie, and DNS at 3 a.m. Everything is written from real operational experience.

More about the project →