Randy Isbell <jisbell@cisco.com> writes:
> One thing I do find interesting: while the backup created with my
> "online" / Chap 23.3 processing contains 18 WAL files, only the last
> 3 (most recent) are used during the recovery process. Is that normal?
Hmm, no, that seems pretty suspicious. The thing *should* be starting
from the checkpoint created by pg_start_backup(), but that observation
makes it sound like it is starting from some later one, probably
whatever happened to be current when pg_control got archived. Please
check this by noting what the startup messages say about where replay
commences, and compare to the backup label file (plain text, you can
just look at it) and pg_control (use pg_controldata before starting the
replay).
regards, tom lane