Neat. Keep in mind though, backing up fly.toml
isn’t going to be enough in some scenarios:
- As part of
deploy
s, changes insecrets
are propogated too. Harder to roll that back. - Unsure how infrastructure related changes driven by
fly.toml
(like mountpoints, say) might behave.
For our setup right now, for rollbacks we keep around 2 Fly apps: One runs the last known good version (say, GOOD
), and one runs the current code (say, LATEST
). If LATEST
is going bust, we switch over the DNS entries to point to GOOD
: Rollback - #15 by ignoramous