postgres in a container on the same box, or managed at $19/mo for 8gb of data Backups

Things that actually went wrong for me, ordered by how much they hurt:

1. Major version upgrade. Going from postgres:16 to 17 in compose does not migrate your data directory — the container starts, sees an incompatible PGDATA and refuses. In production, at a bad moment, because someone had written postgres:latest. Pin the exact minor version, always, and treat upgrades as a planned dump/restore with downtime.
2. Disk full. Postgres does not degrade gracefully, it stops accepting writes. Alert at 70%, not 90%.
3. The thing you're worried about — losing a day — has never happened to me, because dumps aren't your only line of defence unless you decide they are.

Your colleague is right about the exposure and wrong about the fix. You don't need managed to close a 24-hour window, you need WAL archiving. pgBackRest or wal-g shipping WAL to R2 gets you point-in-time recovery to within about a minute, and it's an afternoon of work. That capability is most of what managed is actually selling you.

Postgres in a container is fine. Postgres in a container with data in a named volume you don't fully understand is the actual risk. Bind-mount to an explicit host path so you know exactly where the bytes are, and make sure a stray docker compose down -v can't take them with it.

Also shm_size. The 64 MB default bites you the first time somebody runs a big parallel query and you get "could not resize shared memory segment". Set it to 1g and forget about it.

Container versus managed is slightly the wrong axis. The question is who is on the hook when it's broken, and on a two-person team shipping fast the answer is "whoever least wants to be".

$19 is roughly 40 minutes of one developer's time a month. If self-hosting costs you more than 40 minutes a month averaged over a year — including the one bad night — managed is cheaper. In my experience at 8 GB with simple usage, self-hosting comes in well under that, as long as you do the WAL thing.

The extra network hop is real but small — same region is maybe 1-3 ms per query. What actually hurts is a chatty ORM doing 40 sequential queries per request, where 2 ms becomes 80. Count your queries per request before deciding it doesn't matter.

A local socket to a container on the same box is genuinely faster, and for some workloads noticeably so.

Whatever you choose, separate it from the backup fix. Do WAL archiving this week regardless. If you migrate to managed in three months you've lost nothing and you'll know how your own restore works — which you'll want anyway the day the managed provider has a bad afternoon and you need a copy somewhere else.

8gb fits in ram on a 15 euro box, self host. managed postgres at that size is paying for a checkbox