We had a customer whose instance of OpenNMS stopped abruptly. Turns out their number of events exceeded the ability of PostgreSQL to process since the eventid exceeded 2147483647.

Probably need to change it to bigint or numeric.

Just as a note, 2147483647 corresponds to 68 events/sec over a year. I'm surprised other clients haven't seen this. It'd a big deal, since the only workaround is to reset the index (which I did by blowing away the event data which wasn't important to this client).