All,
So have encountered a 2nd report of this issue, or of an issue which
sounds very similar:
- corruption in two "queue" tables
- the tables are written in a high-concurrency, lock-contested environment
- user uses SELECT FOR UPDATE with these tables.
- pg_stat_statements .so is loaded, but the extension is not installed
- four rows were added to the queue tables, while not being added to the
PK index. This allowed duplicate PKs to be added.
Currently the user is testing not loading the pg_stat_statements.so to
see if the problem goes away. They have a destruction test environment,
so we should be able to confirm/deny in a couple days.
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com