Re: Vacuum analyze infos - Mailing list pgsql-admin

From Tom Lane
Subject Re: Vacuum analyze infos
Date
Msg-id 21457.1031686469@sss.pgh.pa.us
Whole thread Raw
In response to Vacuum analyze infos  (Jean-Arthur Silve <jeanarthur@eurovox.fr>)
List pgsql-admin
Jean-Arthur Silve <jeanarthur@eurovox.fr> writes:
> NOTICE:  Pages 7430: Changed 0, Empty 0; Tup 154392: Vac 6271, Keep 0,
> UnUsed 451218.
>          Total CPU 0.05s/0.34u sec elapsed 1.80 sec.

> what is the difference wetween Vac and and unUsed ?

IIRC, "Vac" is the number of tuples freed in this VACUUM pass.
"UnUsed" is the total number of free tuple slots (linepointers) in the
table.  I'm not sure if that includes the just-freed ones or not.

> Is unUsed spaces unUsed but not deleted ?

Freed slot pointers are available for re-use later.  But the fact that
there are many more free pointers than used pointers (450k vs 155k in
your example) suggests to me that a VACUUM FULL might be appropriate.

If you find that VACUUM FULL reduces the file size (number of pages)
significantly, then you should plan on more frequent regular VACUUMs
and/or increasing the postmaster's FSM parameters to prevent dead
space from accumulating again.

            regards, tom lane

pgsql-admin by date:

Previous
From: Tom Lane
Date:
Subject: Re: fsync or fdatasync
Next
From: Ragnar Kjørstad
Date:
Subject: Re: fsync or fdatasync