A set of functions to handle low-level management of relations using mincore to explore cache memory.
With PostgreSQL, each Table or Index is splitted in segments of (usually) 1GB, and each segment is splitted in pages in memory then in blocks for the filesystem.
Those functions let you know which and how many disk block from a relation are in the page cache of the operating system. It can provide the result as a VarBit and can be stored in a table. Then using this table, it is possible to restore the page cache state for each block of the relation, even in another server, thanks to Streaming Replication.
Other functions are used to set a POSIX_FADVISE flag on the entire relation (each segment). The more usefull are probably WILLNEED and DONTNEED which push and pop blocks of each segments of a relation from page cache, respectively.
Each functions are call with at least a table name or an index name (or oid) as a parameter and walk each segment of the relation.
You can grab the latest code with git:
git clone git://git.postgresql.org/git/pgfincore.git
or
git://github.com/klando/pgfincore.git
And the project is on pgfoundry : http://pgfoundry.org/projects/pgfincore
From source code:
make clean
make
su
make install
For PostgreSQL >= 9.1, log in your database and:
mydb=# CREATE EXTENSION pgfincore;
For other release, create the functions from the sql script (it should be in your contrib directory):
psql mydb -f pgfincore.sql
PgFincore is also shipped with Debian scripts to build your own package:
aptitude install debhelper postgresql-server-dev-all postgresql-server-dev-9.1
# or postgresql-server-dev-8.4|postgresql-server-dev-9.0
make deb
dpkg -i ../postgresql-9.1-pgfincore_1.1.1-1_amd64.deb
PgFincore is packaged for RPM at http://yum.postgresql.org/ PgFincore is packaged for debian at http://pgapt.debian.net/
Here are some examples of usage. If you want more details go to Documentation_
May be useful:
cedric=# select * from pgfincore('pgbench_accounts');
relpath | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit | pages_dirty | group_dirty
--------------------+---------+--------------+--------------+-----------+-----------+---------------+---------+-------------+-------------
base/11874/16447 | 0 | 4096 | 262144 | 262144 | 1 | 81016 | | 0 | 0
base/11874/16447.1 | 1 | 4096 | 65726 | 65726 | 1 | 81016 | | 0 | 0
(2 rows)
Time: 31.563 ms
You may want to try to keep a table or an index into the OS Page Cache, or preload a table before your well know big query is executed (reducing the query time).
To do so, just execute the following query:
cedric=# select * from pgfadvise_willneed('pgbench_accounts');
relpath | os_page_size | rel_os_pages | os_pages_free
--------------------+--------------+--------------+---------------
base/11874/16447 | 4096 | 262144 | 169138
base/11874/16447.1 | 4096 | 65726 | 103352
(2 rows)
Time: 4462,936 ms
- The column os_page_size report that page size is 4KB.
- The column rel_os_pages is the number of pages of the specified file.
- The column os_pages_free is the number of free pages in memory (for caching).
You may want to restore a table or an index into the OS Page Cache as it was while you did the snapshot. For example if you have to reboot your server, then when PostgreSQL start up the first queries might be slower because neither PostgreSQL or the OS have pages in their respective cache about the relations involved in those first queries.
Executing a snapshot and a restore is very simple:
-- Snapshot
cedric=# create table pgfincore_snapshot as
cedric-# select 'pgbench_accounts'::text as relname,*,now() as date_snapshot
cedric-# from pgfincore('pgbench_accounts',true);
-- Restore
cedric=# select * from pgfadvise_loader('pgbench_accounts', 0, true, true,
(select databit from pgfincore_snapshot
where relname='pgbench_accounts' and segment = 0));
relpath | os_page_size | os_pages_free | pages_loaded | pages_unloaded
------------------+--------------+---------------+--------------+----------------
base/11874/16447 | 4096 | 80867 | 262144 | 0
(1 row)
Time: 35.349 ms
- The column pages_loaded report how many pages have been read to memory (they may have already been in memoy)
- The column pages_unloaded report how many pages have been removed from memory (they may not have already been in memoy);
pgsysconf(OUT os_page_size bigint, OUT os_pages_free bigint,
OUT os_total_pages bigint)
RETURNS record
pgsysconf_pretty(OUT os_page_size text, OUT os_pages_free text,
OUT os_total_pages text)
RETURNS record
pgfadvise(IN relname regclass, IN fork text, IN action int,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_willneed(IN relname regclass,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_dontneed(IN relname regclass,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_normal(IN relname regclass,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_sequential(IN relname regclass,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_random(IN relname regclass,
OUT relpath text, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT os_pages_free bigint)
RETURNS setof record
pgfadvise_loader(IN relname regclass, IN fork text, IN segment int,
IN load bool, IN unload bool, IN databit varbit,
OUT relpath text, OUT os_page_size bigint,
OUT os_pages_free bigint, OUT pages_loaded bigint,
OUT pages_unloaded bigint)
RETURNS setof record
pgfadvise_loader(IN relname regclass, IN segment int,
IN load bool, IN unload bool, IN databit varbit,
OUT relpath text, OUT os_page_size bigint,
OUT os_pages_free bigint, OUT pages_loaded bigint,
OUT pages_unloaded bigint)
RETURNS setof record
pgfincore(IN relname regclass, IN fork text, IN getdatabit bool,
OUT relpath text, OUT segment int, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT pages_mem bigint,
OUT group_mem bigint, OUT os_pages_free bigint,
OUT databit varbit, OUT pages_dirty bigint,
OUT group_dirty bigint)
RETURNS setof record
pgfincore(IN relname regclass, IN getdatabit bool,
OUT relpath text, OUT segment int, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT pages_mem bigint,
OUT group_mem bigint, OUT os_pages_free bigint,
OUT databit varbit, OUT pages_dirty bigint,
OUT group_dirty bigint)
RETURNS setof record
pgfincore(IN relname regclass,
OUT relpath text, OUT segment int, OUT os_page_size bigint,
OUT rel_os_pages bigint, OUT pages_mem bigint,
OUT group_mem bigint, OUT os_pages_free bigint,
OUT databit varbit, OUT pages_dirty bigint,
OUT group_dirty bigint)
RETURNS setof record
This function output size of OS blocks, number of free page in the OS Page Buffer.
cedric=# select * from pgsysconf();
os_page_size | os_pages_free | os_total_pages
--------------+---------------+----------------
4096 | 80431 | 4094174
The same as above, but with pretty output.
cedric=# select * from pgsysconf_pretty();
os_page_size | os_pages_free | os_total_pages
--------------+---------------+----------------
4096 bytes | 314 MB | 16 GB
This function set WILLNEED flag on the current relation. It means that the Operating Sytem will try to load as much pages as possible of the relation. Main idea is to preload files on server startup, perhaps using cache hit/miss ratio or most required relations/indexes.
cedric=# select * from pgfadvise_willneed('pgbench_accounts');
relpath | os_page_size | rel_os_pages | os_pages_free
--------------------+--------------+--------------+---------------
base/11874/16447 | 4096 | 262144 | 80650
base/11874/16447.1 | 4096 | 65726 | 80650
This function set DONTNEED flag on the current relation. It means that the Operating System will first unload pages of the file if it need to free some memory. Main idea is to unload files when they are not usefull anymore (instead of perhaps more interesting pages)
cedric=# select * from pgfadvise_dontneed('pgbench_accounts');
relpath | os_page_size | rel_os_pages | os_pages_free
--------------------+--------------+--------------+---------------
base/11874/16447 | 4096 | 262144 | 342071
base/11874/16447.1 | 4096 | 65726 | 408103
This function set NORMAL flag on the current relation.
This function set SEQUENTIAL flag on the current relation.
This function set RANDOM flag on the current relation.
This function allow to interact directly with the Page Cache. It can be used to load and/or unload page from memory based on a varbit representing the map of the pages to load/unload accordingly.
Work with relation pgbench_accounts, segment 0, arbitrary varbit map:
-- Loading and Unloading
cedric=# select * from pgfadvise_loader('pgbench_accounts', 0, true, true, B'111000');
relpath | os_page_size | os_pages_free | pages_loaded | pages_unloaded
------------------+--------------+---------------+--------------+----------------
base/11874/16447 | 4096 | 408376 | 3 | 3
-- Loading
cedric=# select * from pgfadvise_loader('pgbench_accounts', 0, true, false, B'111000');
relpath | os_page_size | os_pages_free | pages_loaded | pages_unloaded
------------------+--------------+---------------+--------------+----------------
base/11874/16447 | 4096 | 408370 | 3 | 0
-- Unloading
cedric=# select * from pgfadvise_loader('pgbench_accounts', 0, false, true, B'111000');
relpath | os_page_size | os_pages_free | pages_loaded | pages_unloaded
------------------+--------------+---------------+--------------+----------------
base/11874/16447 | 4096 | 408370 | 0 | 3
This function provide information about the file system cache (page cache).
cedric=# select * from pgfincore('pgbench_accounts');
relpath | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit | pages_dirty | group_dirty
--------------------+---------+--------------+--------------+-----------+-----------+---------------+---------+-------------+-------------
base/11874/16447 | 0 | 4096 | 262144 | 3 | 1 | 408444 | | 0 | 0
base/11874/16447.1 | 1 | 4096 | 65726 | 0 | 0 | 408444 | | 0 | 0
For the specified relation it returns:
- relpath : the relation path
- segment : the segment number analyzed
- os_page_size : the size of one page
- rel_os_pages : the total number of pages of the relation
- pages_mem : the total number of relation's pages in page cache. (not the shared buffers from PostgreSQL but the OS cache)
- group_mem : the number of groups of adjacent pages_mem
- os_page_free : the number of free page in the OS page cache
- databit : the varbit map of the file, because of its size it is useless to output Use pgfincore('pgbench_accounts',true) to activate it.
- pages_dirty : if HAVE_FINCORE constant is define and the platorm provides the relevant information, like pages_mem but for dirtied pages
- group_dirty : if HAVE_FINCORE constant is define and the platorm provides the relevant information, like group_mem but for dirtied pages
You can debug the PgFincore with the following error level: DEBUG1 and DEBUG5.
For example:
set client_min_messages TO debug1; -- debug5 is only usefull to trace each block
- PgFincore needs mincore() or fincore() and POSIX_FADVISE
-
PgFincore has a limited mode when POSIX_FADVISE is not provided by the platform.
-
PgFincore needs PostgreSQL >= 8.3
-
PgFincore does not work on windows.
Data Bene, PostgreSQL Expertise, Technical Support and Assistance, Trainings: