Re: bloated postgres data folder, clean up

Поиск
Список
Период
Сортировка
От Rémi Cura
Тема Re: bloated postgres data folder, clean up
Дата
Msg-id CAJvUf_vkFWTt8ADXtJSe965aB+n2kqy91NYRsadyJ_aO-x7obA@mail.gmail.com
обсуждение исходный текст
Ответ на Re: bloated postgres data folder, clean up  (Johnny Morano <johnny.morano@payon.com>)
Ответы Re: bloated postgres data folder, clean up  (Julien Rouhaud <julien.rouhaud@dalibo.com>)
Список pgsql-general
Hey,
first I forgot something in the querry to remove the annoying .XXX :
-----------
SELECT distinct substring(file_name from '\d+' )
FROM find_useless_postgres_file('your_database_name') ;
-----------

Now it seems you do everything all right,
with a slight confusion between bloating and useless files.

From what I understood, bloating is when postgres creates lots of file
 as a cache for a table that once was big, or saw a lot of actions.
Still, the created files are referenced by postgres and postgres might use them.
They are _not_ to be deleted, and are _not_ useless.
You can reduce the number by `VACUUM FULL`

On the other hand, you may end up with useless files in your file system
that are _not_ referenced by postgres in any way,
and are basically trash, uselss for anything.
(broken files)
On way to produce those useless files is for example to launch a vacuum full
(create a lot of files, then use those to clean table, then delete those files),
and kill this vacuum full before it ends.
There might by other causes but I'm not an expert.

So to conclude, the fact that my function doesn't find useless file in you postgres database folder is rather good news,
and by no mean imply that your tables are bloated (or not)

Cheers
Rémi-C



2016-03-03 8:31 GMT+01:00 Johnny Morano <johnny.morano@payon.com>:

Hi,

 

So, I still think the SQL function isnt really working well.

Here’s what I did:

 

/data/postgres # psql

 

postgres=# \copy ( select * from find_useless_postgres_file('live') ) to /tmp/useless_files.csv delimiter ';' csv header;

postgres=# \q

 

/data/postgres # wc -l /tmp/useless_files.csv

7422 /tmp/useless_files.csv

 

# filter out the .## files, e.g.:

# 48175847.37;/base/16398/48175847.37;/data/postgres/base/16398/48175847.37;1047420928

# 48175847.36;/base/16398/48175847.36;/data/postgres/base/16398/48175847.36;1073741824

# 48175847.35;/base/16398/48175847.35;/data/postgres/base/16398/48175847.35;1073741824

# 48175847.34;/base/16398/48175847.34;/data/postgres/base/16398/48175847.34;1073741824

# 48175847.33;/base/16398/48175847.33;/data/postgres/base/16398/48175847.33;1073741824

# 48175847.32;/base/16398/48175847.32;/data/postgres/base/16398/48175847.32;1073741824

# 48175847.31;/base/16398/48175847.31;/data/postgres/base/16398/48175847.31;1073741824

# 48175847.30;/base/16398/48175847.30;/data/postgres/base/16398/48175847.30;1073741824

#

# because oid2name doesn't like them, gives error:

# /data/postgres # oid2name -f 48175847.30 -i -S -q -d live

# oid2name: query failed: ERROR:  invalid input syntax for type oid: "48175847.30"

# LINE 11:   (pg_catalog.pg_relation_filenode(c.oid) IN ('48175847.30')...

#                                                        ^

#

# oid2name: query was: SELECT pg_catalog.pg_relation_filenode(c.oid) as "Filenode", relname as "Table Name"

# FROM pg_catalog.pg_class c

#      LEFT JOIN pg_catalog.pg_namespace n ON n.oid = c.relnamespace

#      LEFT JOIN pg_catalog.pg_database d ON d.datname = pg_catalog.current_database(),

#      pg_catalog.pg_tablespace t

# WHERE relkind IN ('r', 'm', 'i', 'S', 't') AND

#            t.oid = CASE

#                   WHEN reltablespace <> 0 THEN reltablespace

#                   ELSE dattablespace

#            END AND

#   (pg_catalog.pg_relation_filenode(c.oid) IN ('48175847.30'))

# ORDER BY relname

 

/data/postgres # cut -d ';' -f1 /tmp/useless_files.csv | cut -d'.' -f1 | sort -n | uniq >/tmp/potential_useless_oids.csv

/data/postgres # wc -l /tmp/potential_useless_oids.csv

1017 /tmp/potential_useless_oids.csv

 

# get a list of all used oids, instead of examing one by one

/data/postgres # oid2name -i -S -q -d live | awk '{print $1}' | sort -n | uniq >/tmp/used_oids.csv

/data/postgres # wc -l /tmp/used_oids.csv

940 /tmp/used_oids.csv

 

/data/postgres # while read i; do grep $i /tmp/used_oids.csv >/dev/null || (echo "$i" >>/tmp/not_in_use_oids.csv); done < /tmp/potential_useless_oids.csv

/data/postgres # wc -l /tmp/not_in_use_oids.csv

168 /tmp/not_in_use_oids.csv

 

/data/postgres # egrep -v 'fsm|vm' /tmp/not_in_use_oids.csv |wc -l

1

/data/postgres # egrep -v 'fsm|vm' /tmp/not_in_use_oids.csv      

file_name

# The CSV header only

 

So, no bloated files.

But, using the function from https://wiki.postgresql.org/wiki/Show_database_bloat I get:

 

live=# select tbloat,wasted_space from table_bloat order by wasted_space desc limit 25;

tbloat │ wasted_space

────────┼──────────────

    1.0 │ 9976 kB

    1.2 │ 98 GB

    1.0 │ 97 MB

    1.4 │ 96 kB

    1.2 │ 920 kB

    1.2 │ 88 kB

    1.1 │ 88 kB

    2.0 │ 8192 bytes

    0.0 │ 8192 bytes

    1.3 │ 8192 bytes

    2.0 │ 8192 bytes

    1.3 │ 8192 bytes

    1.5 │ 8192 bytes

    1.5 │ 8192 bytes

    2.0 │ 8192 bytes

    1.1 │ 8192 bytes

    1.0 │ 8192 bytes

    1.1 │ 8192 bytes

    1.3 │ 8192 bytes

    1.5 │ 8192 bytes

    1.1 │ 80 kB

    1.0 │ 7584 kB

    1.6 │ 71 MB

    1.0 │ 704 kB

    1.1 │ 6968 kB

(25 rows)

 

So actually, quite a lot of bloated data J

What am I doing wrong?

 

Mit freundlichen Grüßen / With kind regards,

Johnny Morano

____________________________________________________

 

Johnny Morano  |  Principal Systems Engineer

 

PAY.ON GmbH  |  AN ACI WORLDWIDE COMPANY  |  WWW.PAYON.COM

Jakob-Haringer-Str. 1  |  5020 Salzburg  |  Austria

 

This email message and any attachments may contain confidential, proprietary or non-public information. This information is intended solely for the designated recipient(s). If an addressing or transmission error has misdirected this email, please notify the sender immediately and destroy this email. Any review, dissemination, use or reliance upon this information by unintended recipients is prohibited. Any opinions expressed in this email are those of the author personally.

 

From: Rémi Cura [mailto:remi.cura@gmail.com]
Sent: Mittwoch, 2. März 2016 17:49
To: Johnny Morano
Cc: Alvaro Herrera; PostgreSQL General


Subject: Re: [GENERAL] bloated postgres data folder, clean up

 

Hey,

this is quite the *opposite*.

The function find files in the postgres database folder that are not used by the database.

To use it :

 * connect to the database you want to analyse ( **mandatory** ).

 * create the function (execute function definition)

 * Execute `SELECT * FROM find_useless_postgres_file('your_database_name')`


This will output a list of files that are on the disk but not used by postgres,

and so can be removed.

To be extra sure, you should use oid2name programme to check that the useless files are really useless.


For this :
 * output the list of potential useless files with copy for instance
  ex :
  COPY ( SELECT file_name

         FROM find_useless_postgres_file('your_database_name')

   ) TO 'path_to_you_database_folder/potential_useless.txt'

   now you've got a file with a list of potential erroneous files.

 * Then use oid2name

  `$su postgres
   $cd path_to_you_database_folder

   $while read i; do oid2name -f "$i" -i -S -q -d your_database_name; done < potential_useless.txt
  `

  Nothing should show, meaning that every potential erroneous file
   has not been recognized by oid2name !

  If you feel unconvinced, you can manually try oid2name on some
   of the potential erroneous files, to be extra sure.
  It should not find anything.

 

 * Now delete all the files in `potential_useless.txt`.

  It could be wiser to not delete the files but rename those

  (for instance, adding `.potentially_useless` as a postfix)

  so if it breaks something, you have an easy way to revert everything.

 

Anyway, use *-*extra extra*-* caution if you delete.
Except a backup, there would be no easy way to correct a mistake.

Cheers,

Rémi-C

 

2016-03-02 15:38 GMT+01:00 Johnny Morano <johnny.morano@payon.com>:

Hi Remi!

 

This SQL function you have provided, seems to return all valid files, is that correct? In my case, it returned all my ‘base/’ files. Is that normal?

If yes, maybe you rename the function to ‘find_useful_postgres_files’ ;-)

 

Could you explain in steps how to use this function to make a cleanup of bloated data? (like in an example with commands and example output, if possible of course)

 

Thanks!

 

 

Mit besten Grüßen / With best regards,

Johnny Morano

____________________________________________________

 

Johnny Morano  | Principal Systems Engineer

 

PAY.ON GmbH  |  AN ACI WORLDWIDE COMPANY  |  WWW.PAYON.COM

Jakob-Haringer-Str. 1  |  5020 Salzburg  |  Austria

Registered at: LG Salzburg  |  Company number: FN 315081 f  |  VAT-ID: ATU64439405

Managing Director: Christian Bamberger

 

 

Follow us on:

 

cid:image001.jpg@01D126D0.E1AB0670  cid:image002.jpg@01D126D0.E1AB0670  cid:image003.jpg@01D126D0.E1AB0670  

 

This email message and any attachments may contain confidential, proprietary or non-public information. This information is intended solely for the designated recipient(s). If an addressing or transmission error has misdirected this email, please notify the sender immediately and destroy this email. Any review, dissemination, use or reliance upon this information by unintended recipients is prohibited. Any opinions expressed in this email are those of the author personally.

 

This message and any attachments have been scanned for viruses prior leaving PAY.ON; however, PAY.ON does not guarantee the security of this message and will not be responsible for any damages arising as a result of any virus being passed on or arising from any alteration of this message by a third party. PAY.ON may monitor e-mails sent to and from PAY.ON.

 

 

 

 

From: pgsql-general-owner@postgresql.org [mailto:pgsql-general-owner@postgresql.org] On Behalf Of Rémi Cura
Sent: Mittwoch, 2. März 2016 14:58
To: Alvaro Herrera
Cc: PostgreSQL General
Subject: Re: [GENERAL] bloated postgres data folder, clean up

 

Would gladly do it,

but still this "wiki cooloff" stuff,

can't create a page

Cheers,

Rémi-C

 

2016-02-29 20:44 GMT+01:00 Alvaro Herrera <alvherre@2ndquadrant.com>:

Rémi Cura wrote:
> Hey dear list,
> after a fex years of experiments and crash,
> I ended up with a grossly bloated postgres folder.
> I had about 8 Go of useless files.

Would you add a new page to the wiki with this?

https://wiki.postgresql.org/wiki/Category:Administrative_Snippets

--
Álvaro Herrera                http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

 

 


Вложения

В списке pgsql-general по дате отправления:

Предыдущее
От: Tom Lane
Дата:
Сообщение: Re: C function migration from 9.2 to 9.5
Следующее
От: Geoff Winkless
Дата:
Сообщение: space required before negative