pgfincore introduces

pgfincore 和shared buffer不同,是OS层面的缓存,可以把大对象缓存到OS的cache里,其实就是内存,所以机器的内存配置越大越好,至于oracle是缓存到buffer_pool_keep,可以手工刷出,不过这个是database的缓存,和pg的shared buffer同理。这里就不再详细介绍。

1.下载wget http://pgfoundry.org/frs/download.php/3186/pgfincore-v1.1.1.tar.gz

2.解压下载的安装包tar -zxvf pgfincore-v1.1.1.tar.gz

3.复制解压后的目录到pg源代码目录/contrib

4.安装:

postgres用户:

make clean
make

su - root

source /home/postgres/.bash_profile

make install

根据READ.rst

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

 

postgres=# select version();
                                                    version                                                    
---------------------------------------------------------------------------------------------------------------
 PostgreSQL 9.1.2 on x86_64-unknown-linux-gnu, compiled by gcc (GCC) 4.1.2 20071124 (Red Hat 4.1.2-42), 64-bit

postgres=# CREATE EXTENSION pgfincore;
CREATE EXTENSION

postgres=# select * from pg_extension ;
    extname     | extowner | extnamespace | extrelocatable | extversion | extconfig | extcondition 
----------------+----------+--------------+----------------+------------+-----------+--------------
 plpgsql        |       10 |           11 | f              | 1.0        |           | 
 pg_buffercache |       10 |         2200 | t              | 1.0        |           | 
 pgfincore      |       10 |         2200 | t              | 1.1.1      |           |

pgfincore使用:

pgfincore 对象的cache情况

pgfadvise_willneed 将对象刷入cache

pgfadvise_dontneed 将对象刷出cache

pgfadvise_loader 直接和page cache交互,加载,卸载页面

pgsysconf 操作系统的cache情况

pgsysconf_pretty同上,只不过输更易懂,带上了单位。

 

postgres=# \d t
                         Table "public.t"
 Column |  Type   |                   Modifiers                   
--------+---------+-----------------------------------------------
 s      | integer | not null default nextval('t_s_seq'::regclass)
 i      | integer | 
 postgres=# insert into t(i) values (generate_series(1,10000000));
INSERT 0 10000000
postgres=#  select pg_size_pretty(pg_relation_size('t'));
 pg_size_pretty 
----------------
 346 MB

当前文件系统的使用情况:pgsysconf(),pgsysconf_pretty()
 postgres=#  select * from pgsysconf();
 os_page_size | os_pages_free | os_total_pages 
--------------+---------------+----------------
         4096 |        182787 |        2044328

postgres=# select * from pgsysconf_pretty();
 os_page_size | os_pages_free | os_total_pages 
--------------+---------------+----------------
 4096 bytes   | 721 MB        | 7986 MB
(1 row)

表t的使用情况:pgfincore

postgres=#  select * from pgfincore('t');
     relpath      | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit 
------------------+---------+--------------+--------------+-----------+-----------+---------------+---------
 base/12699/16441 |       0 |         4096 |        88496 |     88496 |         1 |        181767 |

os_page_size:文件系统页大小

rel_os_pages:占用文件系统页数量

pages_mem:有多少文件系统页在系统cache

 

刷入cache:pgfadvise_willneed()

postgres=# select * from pgfadvise_willneed('t');
     relpath      | os_page_size | rel_os_pages | os_pages_free 
------------------+--------------+--------------+---------------
 base/12699/16441 |         4096 |        88496 |        186428
(1 row)

postgres=#  select * from pgfincore('t');
     relpath      | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit 
------------------+---------+--------------+--------------+-----------+-----------+---------------+---------
 base/12699/16441 |       0 |         4096 |        88496 |     88496 |         1 |        186397 | 
(1 row)

 

 刷出cache:    
postgres=# select * from pgfadvise_dontneed('t');
     relpath      | os_page_size | rel_os_pages | os_pages_free 
------------------+--------------+--------------+---------------
 base/12699/16441 |         4096 |        88496 |        275021
(1 row)

postgres=# select * from pgfincore('t');
     relpath      | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit 
------------------+---------+--------------+--------------+-----------+-----------+---------------+---------
 base/12699/16441 |       0 |         4096 |        88496 |         0 |         0 |        275052 | 
(1 row)

 

pgfadvise_loader:

postgres=# select * from pgfadvise_loader('t', 0, true, true, B'111000');
     relpath      | os_page_size | os_pages_free | pages_loaded | pages_unloaded 
------------------+--------------+---------------+--------------+----------------
 base/12699/16441 |         4096 |        186626 |            3 |              3
(1 row)

loading:

postgres=# select * from pgfadvise_loader('t', 0, true, false, B'111000');
     relpath      | os_page_size | os_pages_free | pages_loaded | pages_unloaded 
------------------+--------------+---------------+--------------+----------------
 base/12699/16441 |         4096 |        186460 |            3 |              0
(1 row)

unloading:

postgres=# select * from pgfadvise_loader('t', 0, false, true, B'111000');
     relpath      | os_page_size | os_pages_free | pages_loaded | pages_unloaded 
------------------+--------------+---------------+--------------+----------------
 base/12699/16441 |         4096 |        186557 |            0 |              3
(1 row)

pgfadvise_NORMAL
pgfadvise_SEQUENTIAL
pgfadvise_RANDOM

这几个可以指定当前内存的属性,正常,顺序,还是随机。

快照与恢复:

做快照:

create table pgfincore_snapshot as
select 't'::text as relname,*,now() as date_snapshot
from pgfincore('t',true);

应用快照:

select * from pgfadvise_loader('t', 0, true, true,
                          (select databit from  pgfincore_snapshot
                           where relname='t' and segment = 0));

     relpath      | os_page_size | os_pages_free | pages_loaded | pages_unloaded 
------------------+--------------+---------------+--------------+----------------
 base/12699/16441 |         4096 |        186259 |        88490 |              6
(1 row)

postgres=#  select * from pgfincore('t');
     relpath      | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit 
------------------+---------+--------------+--------------+-----------+-----------+---------------+---------
 base/12699/16441 |       0 |         4096 |        88496 |     88490 |         2 |        186097 | 
(1 row)

可以看到 88496 和88490,有几个页面是刷出cache,所以不相等

select * from pgfadvise_willneed('t');

postgres=# select * from pgfincore('t');
     relpath      | segment | os_page_size | rel_os_pages | pages_mem | group_mem | os_pages_free | databit 
------------------+---------+--------------+--------------+-----------+-----------+---------------+---------
 base/12699/16441 |       0 |         4096 |        88496 |     88496 |         1 |        186227 | 
(1 row)


限制:
*PgFincore needs mincore() and POSIX_FADVISE.

*PgFincore has a limited mode when POSIX_FADVISE is not provided by the platform.

*PgFincore needs PostgreSQL >= 8.3

你可能感兴趣的:(pgfincore introduces)