Re: pgsql: Add per-index stats information in verbose logs of autovacuum
От | Peter Geoghegan |
---|---|
Тема | Re: pgsql: Add per-index stats information in verbose logs of autovacuum |
Дата | |
Msg-id | CAH2-WzkYgHZzpGOwR14CScJsjaQpvJrEkEfkh_=wGhzLb=yVdQ@mail.gmail.com обсуждение исходный текст |
Ответ на | pgsql: Add per-index stats information in verbose logs of autovacuum (Michael Paquier <michael@paquier.xyz>) |
Ответы |
Re: pgsql: Add per-index stats information in verbose logs of autovacuum
|
Список | pgsql-committers |
On Mon, Mar 22, 2021 at 9:28 PM Michael Paquier <michael@paquier.xyz> wrote: > Add per-index stats information in verbose logs of autovacuum I think that this is very helpful -- thanks for working on it. However, it seems to me that "%u remain" is misleading. I am referring to this line of output: index "dups_idx": pages: 1431 remain, 1343 newly deleted, 1427 currently deleted, 84 reusable It suggests that the index file can shrink. As if the remaining pages are left now that the pages we deleted have been returned to the OS. I think that this same line of output should look like this instead: index "dups_idx": 1431 pages in total, 1343 newly deleted, 1427 currently deleted, 84 reusable FWIW this is roughly the corresponding point of equivalent VACUUM VERBOSE output: INFO: index "dups_idx" now contains 0 row versions in 1431 pages To be clear, I agree that it would not be sensible to make the log_autovacuum_min_duration output 100% consistent with VACUUM VERBOSE. I just think that "%u remain" is misleading. It's just that one detail. -- Peter Geoghegan
В списке pgsql-committers по дате отправления: