PSA: If you're ever involved in a Postgres XID wraparound related emergency, consider using VACUUM's INDEX_CLEANUP option (or the vacuum_index_cleanup table param) - that way VACUUM just does freezing stuff. VACUUM is *much* faster when run this way. Could easily save the day.
Conversation
Replying to
Hey Peter, this is a really good idea; why don’t “autovacuum for wrap-around” vacuums do this behavior automagically?
1
Replying to
I think that they probably should, at least when it's obviously the right thing. is currently working on something much broader than that exact thing, that will make it possible.