just cleaned house in my mailbox and found this email that got buried. You were spot on with the prepared transactions, we had some hung Hibernate threads that were acting up.
Sorry for resurrecting this ages old thread, but wanted to say a big fat *THANKS*!
Tamás András Kálmán <hello@tamaskalman.com> writes: > we have a PostgreSQL 9.3.4 database with multiple large tables, that keep > growing. > Looking at the pg_stat of the front runner table that is currently about 1 > Gb big, it seems, that stale row data is not marked as dead after updates,
Have you checked for uncommitted prepared transactions? If "select * from pg_prepared_xacts" finds anything, that's probably your problem.