Обсуждение: Vacuum full progress

On 5 Sep 2010, at 12:13, Carlos Henrique Reimer wrote:

> Hi,
>
> I need to shrick a table with 102 GB and approximately 380.000.000 rows.

What exactly are you trying to accomplish? You may be saving some space temporarily by running vacuum full and reindex,
butthe database size will probably grow back to its original size quite quickly once it sees some use
(inserts/deletes).

A table that size usually gets partitioned into smaller tables. How to partition your tables depends on how your data
isorganised and behaves, that's a different discussion. You can find documentation here:
http://www.postgresql.org/docs/8.4/interactive/ddl-partitioning.html

> There is a vacuum full running for 13 hours and the only messages a get are:
>
> INFO:  vacuuming "public.posicoes_controles"
> INFO:  "posicoes_controles": found 43960 removable, 394481459 nonremovable row versions in 13308976 pages
> DETAIL:  0 dead row versions cannot be removed yet.
> Nonremovable row versions range from 193 to 217 bytes long.
> There were 136382074 unused item pointers.
> Total free space (including removable row versions) is 27663157952 bytes.
> 2884123 pages are or will become empty, including 0 at the end of the table.
> 4167252 pages containing 27597464344 free bytes are potential move destinations.
> CPU 100.81s/101.45u sec elapsed 3347.95 sec.
> The relfilenode points to the 38611 object and I see that there are 102 38611.n files in the data directory
representingthis table. From this 102 38611.n files only 53 were updated since the vacuum full process started. 
>
> I need to answer these questions:
>
> a) Is there a way to have a general idea if the process is at the end? Is there a way to identify the progess of the
work?

Well, you already noticed it's at the 53rd file of 102 files. It's about half-way then I guess.

> b) How much space will be shrank at the time vacuum full finishes?

According to the above up to 27GB (probably less), not counting the index bloat you're generating by running vacuum
full.

Alban Hertroys

--
Screwing up is an excellent way to attach something to the ceiling.


!DSPAM:737,4c83781710401762263338!

On Sun, Sep 5, 2010 at 5:09 AM, Carlos Henrique Reimer
<carlos.reimer@opendb.com.br> wrote:
> Hi Alban,
>
> The need for the vacuum full is because there were a problem with the daily
> schedulled vacuum analyze and autovacuum regarding the max_fsm_pages. As it
> was underestimated the vacuum process was not able to flag the pages to be
> reused.
>
> I've cancelled the vacuum full and will think another approach. Maybe a
> CLUSTER can do the work. Will start a CLUSTER and see if I can check the
> progress looking the size of the new table relfilenode. It will probably
> have less than 102 GB.

fastest way if you can afford downtime is something like;

select * into new_table from old_table order by pkcol;
alter old_table rename to old_table_bak;
alter new_table rename to old_table;

--
To understand recursion, one must first understand recursion.