"John T. Dow" <john@johntdow.com> writes:
> We did not do pg_start_backup/pg_stop_backup.
Ugh.
> What if we never did pg_start_backup? Will it work anyway?
You could maybe make it work, if you had full_page_writes turned on
and have a continuous series of WAL files extending back to before
the manual filesystem backup was started. What pg_start_backup mainly
does for you is to automate things and make sure there is a well-defined
spot at which a successful replay can be started. It's *not* going to
"just work" without pg_start_backup, though. You'd need to manually
fake up a suitable backup label file, and maybe some other hacking.
Otherwise what's likely to happen is that the recovery goes through
but leaves you with a corrupted database anyway.
If the data is worth this much trouble to you, I'd suggest hiring a
Postgres consultant who's experienced in data recovery.
> Also, it is unclear from the documentation whether you must have
> archived wal files to replay the wal files. There seems to be no way
> to replay them other than with recovery.conf, and recovery.conf seems
> to require a restore_command that actually does something.
You can just have it copy from pg_xlog, if all the files you need are
in pg_xlog. That's a pretty uncommon situation though, so there's not
any special easy case for it.
regards, tom lane