Thread: VACUUM FULL vs dump & restore

     Greetings,
 What advantages I lose, when using dump-truncate-restore (table
or whole DB) instead of performing VACUUM FULL ? In both cases I have no access to data, but first is much faster
(by subjective estimate).

Thank you,

Ilya A. Kovalenko       (mailto:shadow@oganer.net)

Hello Ilya,

you have to check for yourself which method is faster - just test it
with a stopwatch;-)

You have to take care, because when you make VACUUM FULL, then it
vacuums also the system tables, etc. of postgres.

I'm not sure if this is the same way VACUUM goes through all objects,
but I'd make a customized vacuum, which finds out first every object
which should be vacuumed by:

select    relname
from    pg_class

You can filter out not wanted objects through the query or when
processing the "VACUUM FULL [object]" or only "VACUUM [object].

In this way I can decide for myself what I want to vacuum, and what I
will do by dump-truncate-restore.

In many cases a normal VACUUM was even faster then the primitive
dump-truncate-restore process. The bottlneck on a VACUUM is as I saw
from my experience on tables with long strings inside and an amount of
hundreds of millions.

Regards,

Aldor

Ilya A. Kovalenko wrote:
>      Greetings,
> 
>   What advantages I lose, when using dump-truncate-restore (table
> or whole DB) instead of performing VACUUM FULL ?
>   In both cases I have no access to data, but first is much faster
> (by subjective estimate).
> 
> Thank you,
> 
> Ilya A. Kovalenko       (mailto:shadow@oganer.net)
> 
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Have you searched our list archives?
> 
>                http://archives.postgresql.org
> 

Another way how to do it with having access to the data in the same timeis to create a new table, named a little bit
differentlyand do an:
 

insert     into [table]2
select     *
from       [table];

Then switch to the second table.

Then you have to do on the first table the TRUNCATE and DROP.

For getting out which table is the actual one you can create a table
which holds the originate table name and the actual table name. When
using plpgsql you can check the table name before building the queries
and then build them with EXECUTE.

Be aware that you cannot do:

SELECT    col1, col2
FROM    gettablename('[table]');

Also be aware to switch back when you do the process again, so you dump
the data from the [table]2 to [table].

For my experience this way was faster then dump-truncate-restore on the
table.

Regards,

Aldor

Ilya A. Kovalenko wrote:
>      Greetings,
> 
>   What advantages I lose, when using dump-truncate-restore (table
> or whole DB) instead of performing VACUUM FULL ?
>   In both cases I have no access to data, but first is much faster
> (by subjective estimate).
> 
> Thank you,
> 
> Ilya A. Kovalenko       (mailto:shadow@oganer.net)
> 
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Have you searched our list archives?
> 
>                http://archives.postgresql.org
>