Alexander Shutyaev <shutyaev@gmail.com> wrote:
> We have the following issue. When we use vacuumdb (NOT full) on
> our postgres database (~320Gb) it takes up ~10Gb of disk space
> which is never returned. Why is the space not returned?
Does that happen every time? (i.e., if you run vacuumdb 10 times
in a row while there is no other activity against the cluster, does
it take up 100GB more space?) If not, you are probably doing an
initial build of information on tuple visibility, free space in the
tables, and column statistics. The cluster can run without these,
but it won't be as efficient. To minimize down time, this data is
built up opportunistically once the service is up and running.
--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company