I forgot to mention that my deployments include automated migrations as often as possible, sometimes destructive for refactoring purpose, as such, to maintain PostgreSQL on a basic linux box I am:
- for having an automated backup prior in the automated deployment script that may play destructive migrations,
- against the needless overhead of coupling both binary and relational data in operations that slows the whole thing down or makes it less reliable
Also got supposedly many new points against, mixed with more detail on the points briefly exposed in my previous email, going deeper in detail, about how it fits in the big picture of my personal practice ... and how this has destabilized my prod for months:
tl;dr
If you store media files in PostgreSQL on a production server, then do take disk space alarms seriously even if they happen only during backups.
Otherwise I fail to see how to avoid a pattern of recurring incidents, "manually unblocking automated deployments" (double debt interest cost because also defeats the purpose of automating deployment), when not filling up a disk during the nightly backup dump ...
Hope this helps,
Have a great day
