we should consider deploying the PostgreSQL database server ~~and Prometheus monitoring~~ (done) from our existing puppet recipes instead of using the omnibus versions. pgsql, in particular, is causing us troubles because our normal backup procedures do not work. ~~prometheus is also causing issues because we cannot reuse our export/realize resources in puppet (see https://gitlab.torproject.org/tpo/tpa/services/-/issues/34420, https://gitlab.torproject.org/tpo/tpa/services/-/issues/33921)~~ fixed