Обсуждение: Postgres Database Disk Usage

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

--

Thank you for the response Vijaykumar,

>Do you have replication active?
No.

>What is the Size of wal /pg_xlog directory ?
[postgres@localhost pg_xlog]$ du --si
4.1k./archive_status
403M.

Let me know if you you need any more information.

--
Sincerely,

Jeremiah


From: Vijaykumar Jain <vjain@opentable.com>
Sent: Friday, February 8, 2019 10:51 AM
To: Jeremiah Bauer
Cc: pgsql-admin@lists.postgresql.org
Subject: Re: [External] Postgres Database Disk Usage

Do you have replication active?
is the size on the replica same as primary?
no stale orphan replication slots?
What is the Size of wal /pg_xlog directory ?

On Fri, 8 Feb 2019 at 9:17 PM Jeremiah Bauer <jbauer@agristats.com> wrote:
Hello,

We have a development database that reports a larger size than the sum of it's relations and I was hoping someone had
someinsight into this issue.  The consumed space is reflected in the size of the data directories PostgreSQL is
maintainingon disk, so it's just not a reported size issue. 

This is the version of Postgres we are using:

FVDM=# select version();
                                                  version
-----------------------------------------------------------------------------------------------------------
 PostgreSQL 9.6.11 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23), 64-bit

pg_database_size reports this size:

    name    |  owner   |  size
------------+----------+---------
 FVDM       | postgres | 43 GB

Running a query to report the total relation size of each schema results in this:

     schemaname     | pg_size_pretty
--------------------+----------------
 information_schema | 152 kB
 pg_catalog         | 7880 kB
 public             | 52 MB
 mart               | 439 MB
(4 rows)

The total size of the relations is less than a gigabyte, but the database is 43GB in size.  I've checked each table and
therearen't any added indexes, so I'm very confused. 

We've run vacuum full and reindex on the database with no reduction in size and a vacuum analyze verbose reports no
recoverablerows.  I cloned the machine for troubleshooting and even attempted truncating every table in the database to
seeif that remedied the issue.  The only thing that has reclaimed the space is dropping the database and restoring from
backup.

What is using the space and how do we reclaim it?

Any help or insight would be greatly appreciated, I'm at a loss as to what is consuming this space.

--


Jeremiah

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential
andprotected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any
disclosure,copying, distribution or use of this email or any attachment is strictly prohibited and you should not read
themessage or read or open any attachment. If you have received this email by mistake, please immediately notify the
senderand delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any
personor entity resulting from the unintended or unauthorized use of any information contained in this email. 
--

Regards,
Vijay
CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential
andprotected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any
disclosure,copying, distribution or use of this email or any attachment is strictly prohibited and you should not read
themessage or read or open any attachment. If you have received this email by mistake, please immediately notify the
senderand delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any
personor entity resulting from the unintended or unauthorized use of any information contained in this email. 

--

---

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

--

Would you use pg_total_relation_size, not pg_relation_size in your sizing query?  This will capture associated indexes and such.

Regards,
Michael Vitale

Vijaykumar Jain

Friday, February 8, 2019 12:05 PM

Ok I am out of ideas :)

Sorry, I’ll let the experts step in here :)

 btw the 43GB and 439MB look close :) I hope there is no conversion error :) 

--

Regards,

Vijay

Jeremiah Bauer

Friday, February 8, 2019 11:29 AM

P {margin-top:0;margin-bottom:0;}

We don't have any blobs or json objects in this database.  I've sanitized the output of the query below, and toast doesn't show up until well down the list.

FVDM=# SELECT nspname || '.' || relname AS "relation",

FVDM-#     pg_size_pretty(pg_relation_size(C.oid)) AS "size"

FVDM-#   FROM pg_class C

FVDM-#   LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)

FVDM-#   WHERE nspname NOT IN ('pg_catalog', 'information_schema')

FVDM-#   ORDER BY pg_relation_size(C.oid) DESC

FVDM-#   LIMIT 20;

                       relation                       |  size  

------------------------------------------------------+---------

 mart.*****_data                                        | 301 MB

 mart.detailed_*****                 | 66 MB

 mart.major_*****              | 58 MB

 public.temp_*****                                 | 20 MB

 public.temp_*****                      | 13 MB

 mart.customer_*****       | 11 MB

 public.temp_*****| 11 MB

 mart.major_*****                  | 1888 kB

 public.temp_*****            | 1880 kB

 public.hold_*****           | 1424 kB

 public.temp_*****                          | 1208 kB

 mart.ranking_*****                                 | 1144 kB

 public.temp_*****          | 1080 kB

 public.temp_*****       | 984 kB

 mart.customer_*****                      | 696 kB

 mart.economic_*****                      | 456 kB

 public.hold_*****                          | 448 kB

 mart.expanded_*****             | 384 kB

 pg_toast.pg_toast_2618                               | 376 kB

 public.temp_*****       | 232 kB

I also queried just for toast tables:

FVDM=# SELECT nspname || '.' || relname AS "relation",

    pg_size_pretty(pg_relation_size(C.oid)) AS "size"

  FROM pg_class C

  LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)

  WHERE nspname NOT IN ('pg_catalog', 'information_schema') and c.relname ~* 'toast'

  ORDER BY pg_relation_size(C.oid) DESC

  LIMIT 20;

           relation            |    size    

-------------------------------+------------

 pg_toast.pg_toast_2618        | 376 kB

 pg_toast.pg_toast_2619        | 48 kB

 pg_toast.pg_toast_2618_index  | 16 kB

 pg_toast.pg_toast_2619_index  | 16 kB

 pg_toast.pg_toast_13104_index | 8192 bytes

 pg_toast.pg_toast_3596_index  | 8192 bytes

 pg_toast.pg_toast_2606_index  | 8192 bytes

 pg_toast.pg_toast_2620_index  | 8192 bytes

 pg_toast.pg_toast_2609_index  | 8192 bytes

 pg_toast.pg_toast_2396_index  | 8192 bytes

 pg_toast.pg_toast_3592_index  | 8192 bytes

 pg_toast.pg_toast_1255_index  | 8192 bytes

 pg_toast.pg_toast_13109_index | 8192 bytes

 pg_toast.pg_toast_13114_index | 8192 bytes

 pg_toast.pg_toast_13119_index | 8192 bytes

 pg_toast.pg_toast_13124_index | 8192 bytes

 pg_toast.pg_toast_13129_index | 8192 bytes

 pg_toast.pg_toast_13134_index | 8192 bytes

 pg_toast.pg_toast_2604_index  | 8192 bytes

 pg_toast.pg_toast_2964_index  | 8192 bytes

--

Jeremiah

---

From: Vijaykumar Jain <vjain@opentable.com>
Sent: Friday, February 8, 2019 11:21 AM
To: Jeremiah Bauer
Cc: pgsql-admin@lists.postgresql.org
Subject: Re: [External] Postgres Database Disk Usage

 

Do you have any blobs or large json objects as relnames?

they generally end up as toast tables .

Which may be the ones holding the space.

On Fri, 8 Feb 2019 at 9:25 PM Jeremiah Bauer <jbauer@agristats.com> wrote:

--

Regards,

Vijay

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

---

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

---

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

P {margin-top:0;margin-bottom:0;}

Sure, I tried that before I posted to the mailing list.

This is the size of the restored database with a different name:

[postgres@localhost ~]$ pg_dump FVDM > /tmp/out.sql

[postgres@localhost ~]$ less /tmp/out.sql

[postgres@localhost ~]$ createdb FVDM_restored

[postgres@localhost ~]$ psql FVDM_restored </tmp/out.sql

....completes no errors....

postgres=# SELECT d.datname AS Name,  pg_catalog.pg_get_userbyid(d.datdba) AS Owner,

    CASE WHEN pg_catalog.has_database_privilege(d.datname, 'CONNECT')

        THEN pg_catalog.pg_size_pretty(pg_catalog.pg_database_size(d.datname))

        ELSE 'No Access'

    END AS SIZE                                                                    

FROM pg_catalog.pg_database d

    ORDER BY

    CASE WHEN pg_catalog.has_database_privilege(d.datname, 'CONNECT')

        THEN pg_catalog.pg_database_size(d.datname)

        ELSE NULL

    END DESC -- nulls first

    LIMIT 20;

     name      |  owner   |  size  

---------------+----------+---------

 FVDM          | postgres | 43 GB

 FVDM_restored | postgres | 497 MB

If you have enough disk space, try to VACUUM FULL the database (or do it one table at a time).

--

Jeremiah

---

From: Mariel Cherkassky <mariel.cherkassky@gmail.com>
Sent: Friday, February 8, 2019 12:42 PM
To: MichaelDBA
Cc: Vijaykumar Jain; Jeremiah Bauer; pgsql-admin@lists.postgresql.org
Subject: [Junk released by User action] Re: [External] Postgres Database Disk Usage

 

Can you try to dump the database and then restore it with a different name ? Check if the size of the restored one is the same. 

On Fri, Feb 8, 2019, 7:12 PM MichaelDBA <MichaelDBA@sqlexec.com wrote:

Would you use pg_total_relation_size, not pg_relation_size in your sizing query?  This will capture associated indexes and such.

Regards,
Michael Vitale

Vijaykumar Jain

Friday, February 8, 2019 12:05 PM

Ok I am out of ideas :)

Sorry, I’ll let the experts step in here :)

 btw the 43GB and 439MB look close :) I hope there is no conversion error :) 

--

Regards,

Vijay

Jeremiah Bauer

Friday, February 8, 2019 11:29 AM

We don't have any blobs or json objects in this database.  I've sanitized the output of the query below, and toast doesn't show up until well down the list.

FVDM=# SELECT nspname || '.' || relname AS "relation",

FVDM-#     pg_size_pretty(pg_relation_size(C.oid)) AS "size"

FVDM-#   FROM pg_class C

FVDM-#   LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)

FVDM-#   WHERE nspname NOT IN ('pg_catalog', 'information_schema')

FVDM-#   ORDER BY pg_relation_size(C.oid) DESC

FVDM-#   LIMIT 20;

                       relation                       |  size  

------------------------------------------------------+---------

 mart.*****_data                                        | 301 MB

 mart.detailed_*****                 | 66 MB

 mart.major_*****              | 58 MB

 public.temp_*****                                 | 20 MB

 public.temp_*****                      | 13 MB

 mart.customer_*****       | 11 MB

 public.temp_*****| 11 MB

 mart.major_*****                  | 1888 kB

 public.temp_*****            | 1880 kB

 public.hold_*****           | 1424 kB

 public.temp_*****                          | 1208 kB

 mart.ranking_*****                                 | 1144 kB

 public.temp_*****          | 1080 kB

 public.temp_*****       | 984 kB

 mart.customer_*****                      | 696 kB

 mart.economic_*****                      | 456 kB

 public.hold_*****                          | 448 kB

 mart.expanded_*****             | 384 kB

 pg_toast.pg_toast_2618                               | 376 kB

 public.temp_*****       | 232 kB

I also queried just for toast tables:

FVDM=# SELECT nspname || '.' || relname AS "relation",

    pg_size_pretty(pg_relation_size(C.oid)) AS "size"

  FROM pg_class C

  LEFT JOIN pg_namespace N ON (N.oid = C.relnamespace)

  WHERE nspname NOT IN ('pg_catalog', 'information_schema') and c.relname ~* 'toast'

  ORDER BY pg_relation_size(C.oid) DESC

  LIMIT 20;

           relation            |    size    

-------------------------------+------------

 pg_toast.pg_toast_2618        | 376 kB

 pg_toast.pg_toast_2619        | 48 kB

 pg_toast.pg_toast_2618_index  | 16 kB

 pg_toast.pg_toast_2619_index  | 16 kB

 pg_toast.pg_toast_13104_index | 8192 bytes

 pg_toast.pg_toast_3596_index  | 8192 bytes

 pg_toast.pg_toast_2606_index  | 8192 bytes

 pg_toast.pg_toast_2620_index  | 8192 bytes

 pg_toast.pg_toast_2609_index  | 8192 bytes

 pg_toast.pg_toast_2396_index  | 8192 bytes

 pg_toast.pg_toast_3592_index  | 8192 bytes

 pg_toast.pg_toast_1255_index  | 8192 bytes

 pg_toast.pg_toast_13109_index | 8192 bytes

 pg_toast.pg_toast_13114_index | 8192 bytes

 pg_toast.pg_toast_13119_index | 8192 bytes

 pg_toast.pg_toast_13124_index | 8192 bytes

 pg_toast.pg_toast_13129_index | 8192 bytes

 pg_toast.pg_toast_13134_index | 8192 bytes

 pg_toast.pg_toast_2604_index  | 8192 bytes

 pg_toast.pg_toast_2964_index  | 8192 bytes

--

Jeremiah

---

From: Vijaykumar Jain <vjain@opentable.com>
Sent: Friday, February 8, 2019 11:21 AM
To: Jeremiah Bauer
Cc: pgsql-admin@lists.postgresql.org
Subject: Re: [External] Postgres Database Disk Usage

 

Do you have any blobs or large json objects as relnames?

they generally end up as toast tables .

Which may be the ones holding the space.

On Fri, 8 Feb 2019 at 9:25 PM Jeremiah Bauer <jbauer@agristats.com> wrote:

--

Regards,

Vijay

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

Do you character varying columns that defined with limit ? Search for pg_toast on google and find the query to determine how much pg_toast you have.

Hello,

We have a development database that reports a larger size than the sum of it's relations and I was hoping someone had some insight into this issue.  The consumed space is reflected in the size of the data directories PostgreSQL is maintaining on disk, so it's just not a reported size issue.

This is the version of Postgres we are using:

FVDM=# select version();

                                                  version                                                  

-----------------------------------------------------------------------------------------------------------

 PostgreSQL 9.6.11 on x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-23), 64-bit

pg_database_size reports this size:

    name    |  owner   |  size  

------------+----------+---------

 FVDM       | postgres | 43 GB

Running a query to report the total relation size of each schema results in this:

     schemaname     | pg_size_pretty

--------------------+----------------

 information_schema | 152 kB

 pg_catalog         | 7880 kB

 public             | 52 MB

 mart               | 439 MB

(4 rows)

The total size of the relations is less than a gigabyte, but the database is 43GB in size.  I've checked each table and there aren't any added indexes, so I'm very confused.

We've run vacuum full and reindex on the database with no reduction in size and a vacuum analyze verbose reports no recoverable rows.  I cloned the machine for troubleshooting and even attempted truncating every table in the database to see if that remedied the issue.  The only thing that has reclaimed the space is dropping the database and restoring from backup.

What is using the space and how do we reclaim it?  

Any help or insight would be greatly appreciated, I'm at a loss as to what is consuming this space.

--

Jeremiah

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

---

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

Since the underlying filesystem shows the 43GB, try to isolate it to what file(s) are causing the problem.  Assuming you are connected to the "FVDM" database and all the tables are in one schema, run this query to identify all the table files.

select c.oid, c.relname, c.reltuples, c.relfilenode,  pg_relation_filenode(c.oid), pg_relation_filepath(c.oid) from pg_class c, pg_tables t where c.relname = t.tablename and t.schemaname = 'myschema' order by 1;

Before drilling down to where the database files are, do a du -hs at $PGDATA directory.  Then do it at $PGDATA/base.  Then do it at the table directory level:
$PGDATA/base/ <database filenode> and run "du -hs" on everything there and see if something stands out.

Jeremiah Bauer

Friday, February 8, 2019 2:59 PM

P {margin-top:0;margin-bottom:0;}

Yes, it did run out of free space when some test data was being loaded.

I added more space, could this be the cause of the issue? I ran a Vacuum Full on the entire database after adding more free space, but the size of the database is still significantly larger than the sum of the objects in it.

--

Jeremiah

---

From: Mariel Cherkassky <mariel.cherkassky@gmail.com>
Sent: Friday, February 8, 2019 2:33 PM
To: Jeremiah Bauer
Cc: MichaelDBA; Vijaykumar Jain; pgsql-admin@lists.postgresql.org
Subject: [Junk released by User action] Re: [Junk released by User action] Re: [External] Postgres Database Disk Usage

 

Is there a chance that you run out of disks space recently? 

On Fri, Feb 8, 2019, 8:34 PM Jeremiah Bauer <jbauer@agristats.com wrote:

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

Jeremiah Bauer

Friday, February 8, 2019 1:34 PM

P {margin-top:0;margin-bottom:0;}

Sure, I tried that before I posted to the mailing list.

This is the size of the restored database with a different name:

[postgres@localhost ~]$ pg_dump FVDM > /tmp/out.sql

[postgres@localhost ~]$ less /tmp/out.sql

[postgres@localhost ~]$ createdb FVDM_restored

[postgres@localhost ~]$ psql FVDM_restored </tmp/out.sql

....completes no errors....

postgres=# SELECT d.datname AS Name,  pg_catalog.pg_get_userbyid(d.datdba) AS Owner,

    CASE WHEN pg_catalog.has_database_privilege(d.datname, 'CONNECT')

        THEN pg_catalog.pg_size_pretty(pg_catalog.pg_database_size(d.datname))

        ELSE 'No Access'

    END AS SIZE                                                                    

FROM pg_catalog.pg_database d

    ORDER BY

    CASE WHEN pg_catalog.has_database_privilege(d.datname, 'CONNECT')

        THEN pg_catalog.pg_database_size(d.datname)

        ELSE NULL

    END DESC -- nulls first

    LIMIT 20;

     name      |  owner   |  size  

---------------+----------+---------

 FVDM          | postgres | 43 GB

 FVDM_restored | postgres | 497 MB

--

Jeremiah

---

From: Mariel Cherkassky <mariel.cherkassky@gmail.com>
Sent: Friday, February 8, 2019 12:42 PM
To: MichaelDBA
Cc: Vijaykumar Jain; Jeremiah Bauer; pgsql-admin@lists.postgresql.org
Subject: [Junk released by User action] Re: [External] Postgres Database Disk Usage

 

Can you try to dump the database and then restore it with a different name ? Check if the size of the restored one is the same. 

On Fri, Feb 8, 2019, 7:12 PM MichaelDBA <MichaelDBA@sqlexec.com wrote:

CONFIDENTIALITY NOTICE: The information contained in this email (and any attachments) is privileged and confidential and protected from disclosure. If you are not the intended recipient of this email or the attachments, be aware that any disclosure, copying, distribution or use of this email or any attachment is strictly prohibited and you should not read the message or read or open any attachment. If you have received this email by mistake, please immediately notify the sender and delete it permanently from your system. Agri Stats, Inc. and its subsidiaries will not be held liable to any person or entity resulting from the unintended or unauthorized use of any information contained in this email.

Mariel Cherkassky

Friday, February 8, 2019 12:42 PM

Can you try to dump the database and then restore it with a different name ? Check if the size of the restored one is the same.