hi,
in our test enviroment, if one database's have major update operations, autovacuum does not work and cause major performance degradation. if found this issue may be resolved by revert this Skip redundant anti-wraparound vacuums ยท postgres/postgres@2aa6e33 (github.com) commit. after fetch some disccusion about this revert, i have some question as follow: 1. i understand that anti-wraparound and no-aggressive autovacuum will be skipped for shared catalog tables, but why this can trigger autovacuum does not work for others tables ? 2. "this could cause autovacuum to lock down", this lock down implict that autovacuum can make a dead lock problem ? 3. how to reproduce this lock down or autovacuum invalid issue, must be cluster enviroment ? so is there any body know these issuse or commits can give me some suggestion about my confusion. | | jiye | | jiye...@126.com |
