Thanks for the clarification. Also, I was wrong about the temp_* columns in the pg_stat_database view in that they report the overall temp file usage since the last stats reset and not the “current” usage.
Is there a known issue with temp file reporting? I have a production db that is reporting 231TB of temp space usage, which can’t be true. temp_blks_written in pg_stat_statements sort of confirms this, as the reported queries show just a few MB of temp space usage. I suspect a reboot will clear this (to do) and wanted to check with you to see if someone else has had similar experiences.
Just to clarify, "temprary files" and "temp blocks" refer to different entities in this context. The "temprary files" are ephemeral storage that is created and alive only for tasks like tuple sorting, whereas "temp blocks" are referring to the storage for temprary tables.
If you're performing a massive insert into a temporary table, you'll find that pg_stat_database.temp_bytes doesn't increase at all.
regards.
-- Kyotaro Horiguchi NTT Open Source Software Center
