Federation is just scraping: the central prometheus server scrapes the other's /federate endpoint.
Scraping only collects metrics.
Alerts are completely separate:
- alerting rules are evaluated periodically inside the prometheus server
- if they are triggered, then alerts are *pushed* out to alertmanager (this is not scraping).
If you want to duplicate this centrally, then you'll need to duplicate the alerting rules on the central server. But why would you want to? The alerts are already coming from the remote prometheus servers, which can be configured to send them to a central alertmanager (which can be a cluster).
If the problem you are trying to solve is what happens when the remote prometheus servers lose connectivity (and hence can't deliver alerts), then you won't be able to scrape the federation endpoint either.