Off AWS eu-west-1, onto Elastx and Hetzner
Eleven services, three managed databases and a queue nobody had documented. We audited for six weeks, rewrote the estate as NixOS configurations, replaced the managed pieces with PostgreSQL and MinIO, and cut over one service a week with rehearsed rollbacks. The queue turned out to be load-bearing; finding that out cost us a fortnight we had not quoted.
- −61%
- Monthly infrastructure spend
- 0
- Unplanned minutes of downtime during cut-over
- 14 wks
- Audit to final decommission