From: Masahiko Sawada [mailto:sawada.m...@gmail.com] > Thank you for suggestion. It sounds more smarter. So it would be more better > if we vacuums database for anti-wraparound in ascending order of > relfrozenxid?
I thought so, too. The current behavior is inconsistent: the launcher tries to assign all workers to one database with the biggest wraparound risk in order to eliminate the risk as fast as possible, but the workers don't get hurry to reduce the risk. Regards Takayuki Tsunakawa