Skip to content
All work

Infrastructure · 2026

Infrastructure Console

One screen for nine services that used to need nine tabs.

A discovery-based operator console across a three-node cluster: containers, hosts, storage, networking, DNS and certificates, with an aggregated feed of everything currently wrong.

The design rule was that it must discover what it monitors rather than hold a hardcoded list. Two separate outages had gone unnoticed for hours precisely because a checker was watching a list that reality had moved past.

It talks to lightweight agents on each host rather than a central management product, which keeps the blast radius small and the dependencies boring.

The console overview: fleet status across four hosts, container health grouped by stack, storage capacity, and network and tunnel status, with open issues called out at the top.
The console overview: fleet status across four hosts, container health grouped by stack, storage capacity, and network and tunnel status, with open issues called out at the top.