Обсуждение: Question on vacuum output

Поиск
Список
Период
Сортировка

Question on vacuum output

От
"Chris Hoover"
Дата:
I am having some performance issues on some of my larger databases (50-100+GB).  I just ran a vacuum analyze on our key table, and saw this at the end.

DETAIL:  562499 dead row versions cannot be removed yet.
There were 15117 unused item pointers.
0 pages are entirely empty.
CPU 8.32s/2.59u sec elapsed 340.19 sec.
INFO:  analyzing "public.clmhdr"
INFO:  "clmhdr": scanned 30000 of 221695 pages, containing 259684 live rows and 75400 dead rows; 30000 rows in sample, 1919021 estimated total rows
VACUUM


What does this mean that there are 562499 dead row versions that could not be removed?  Could this be part of the performance issue?  I have a query that before the vacuum was running in 1-2 seconds and normally after a vacuum analyze will run in a couple hundred miliseconds.  (Although this run of the vacuum analyze does not appear to have helped much, my tell tale query is still running 1sec+).

Thanks,

Chris

Re: Question on vacuum output

От
Alvaro Herrera
Дата:
Chris Hoover wrote:
> I am having some performance issues on some of my larger databases
> (50-100+GB).  I just ran a vacuum analyze on our key table, and saw this at
> the end.
>
> DETAIL:  562499 dead row versions cannot be removed yet.

Is there an open, idle transaction on this database?  A leftover psql
session running in a forgotten, detached screen session maybe?

--
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support