Build a Postgres primary and streaming replica, kill the leader and watch history fork, see why the obvious fix is worse than failing, and recover the old leader with pg_rewind.
Loading content...