2026–W04 #
Thinking about Patroni for PostgreSQL and about how to do failovers between multiple providers in a post-AWS world (where a region is really a colocation provider). Thinking about Wireguard mesh networks between database nodes, or anycasted services. Thinking about how (msnsick) Kubernetes might be a good potential tool for this kind of thing. Thinking about failure, basically.
Also thinking about selfci -- it's very much a thing that I wanted and was going to build for myself but now it exists so I don't have to. (Update: I have played with it, and I like it)
I spent a whole evening tinkering with putting ridiculous PCIe cards on my Chromebook. At one point I thought it might be a lost cause -- I learned that CNVi ports exist which are only compatible with (Intel) WiFi cards and don't run a real PCIe lane, but I've been able to make exactly one PCIe device (already in M.2 form factor) work -- so I know it's possible. I now suspect the particular adaptor I've been using.
I've also been working on PostgreSQL replication at work for zeroish-downtime cut-overs (tm) and now I'm thinking about / planning an architecture document about how to do HA across providers -- in service of a more indie-hosting world and less Amazonication.
Also I just bought four Chromebooks to make either a cluster or (more likely) a hosting environment out of them? This is a bad idea but I'm smiling at it.