■PostgreSQLの統計情報の記録と監視

■パフォーマンス統計情報を取得・保存するPgPerfパッケージ

• 内部のシステムビューなどのほぼ同じ形式で、
• 稼働しているシステムに極力手を加えずに、
• すべてを網羅的に取得・蓄積し、
• 後から自由に分析できるようにする

• スクリプト（SQL、PL/pgSQL）のみで動作するため、PostgreSQLの稼働しているプラットフォームに依存しない。
• 各種性能情報を容易に取得・保存することができ、蓄積したデータを自由に分析・活用することができる。
• インストールおよびアンインストールが容易で、稼働しているPostgreSQLの設定を変更する必要がない。

■PgPerfパッケージのインストール

[snaga@devsv02 pgperf-snapshot]$ psql -f pgperf_install91.sql testdb
BEGIN
CREATE SCHEMA
psql:pgperf_install91.sql:28: NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "snapshot_pkey" for table "snapshot"
CREATE TABLE
CREATE FUNCTION
(...snip...)
CREATE FUNCTION
COMMIT
[snaga@devsv02 pgperf-snapshot]$

[snaga@devsv02 pgperf-snapshot]$ psql testdb
psql (9.0.6, server 9.1.2)
Type "help" for help.

testdb=# SET search_path TO pgperf;
SET
testdb=# \d
                        List of relations
 Schema |                Name                 | Type  |  Owner
--------+-------------------------------------+-------+----------
 pgperf | snapshot                            | table | postgres
 pgperf | snapshot_pg_current_xlog            | table | postgres
 pgperf | snapshot_pg_locks                   | table | postgres
 pgperf | snapshot_pg_relation_size           | table | postgres
 pgperf | snapshot_pg_stat_activity           | table | postgres
 pgperf | snapshot_pg_stat_bgwriter           | table | postgres
 pgperf | snapshot_pg_stat_database           | table | postgres
 pgperf | snapshot_pg_stat_database_conflicts | table | postgres
 pgperf | snapshot_pg_stat_replication        | table | postgres
 pgperf | snapshot_pg_stat_statements         | table | postgres
 pgperf | snapshot_pg_stat_user_indexes       | table | postgres
 pgperf | snapshot_pg_stat_user_tables        | table | postgres
 pgperf | snapshot_pg_statio_user_indexes     | table | postgres
 pgperf | snapshot_pg_statio_user_tables      | table | postgres
 pgperf | snapshot_pg_statistic               | table | postgres
 pgperf | snapshot_pgstatindex                | table | postgres
 pgperf | snapshot_pgstattuple                | table | postgres
(17 rows)

testdb=#

■スナップショットの取得

testdb=# SELECT pgperf.create_snapshot(1);
 create_snapshot
-----------------
               0
(1 row)

testdb=#

■取得したスナップショットの確認

testdb=# SELECT * FROM pgperf.snapshot;
 sid |             ts
-----+----------------------------
   0 | 2012-12-08 11:17:02.822046
(1 row)

testdb=# 

testdb=# SELECT * FROM pgperf.snapshot_pg_stat_bgwriter ;
-[ RECORD 1 ]---------+------------------------------
sid                   | 0
checkpoints_timed     | 136
checkpoints_req       | 35
buffers_checkpoint    | 1746118
buffers_clean         | 0
maxwritten_clean      | 0
buffers_backend       | 1383900
buffers_backend_fsync | 0
buffers_alloc         | 451660
stats_reset           | 2012-11-30 13:29:50.000083+09

testdb=#

■スナップショットの削除

testdb=# SELECT * FROM pgperf.snapshot;
 sid |             ts
-----+----------------------------
   0 | 2012-12-08 11:17:02.822046
(1 row)

testdb=# SELECT pgperf.delete_snapshot(0);
 delete_snapshot
-----------------
 t
(1 row)

testdb=#

■スナップショットの消し込み

testdb=# SELECT pgperf.purge_snapshots('32 days');
 purge_snapshots
-----------------
              12
(1 row)

testdb=#

■PgPerfパッケージのアンインストール

[snaga@devsv02 pgperf-snapshot]$ psql -U postgres -f pgperf_uninstall.sql testdb
psql:pgperf_uninstall.sql:1: NOTICE:  drop cascades to 50 other objects
DETAIL:  drop cascades to table pgperf.snapshot
drop cascades to function pgperf._get_server_version()
drop cascades to function pgperf._check_function(name)
(...snip...)
drop cascades to function pgperf.create_snapshot_pgstatindex(integer)
drop cascades to function pgperf.delete_snapshot_pgstatindex(integer)
DROP SCHEMA
[snaga@devsv02 pgperf-snapshot]$

■まとめ

■使用するスナップショットテーブル

• pgperf.snapshot
• pgperf.snapshot_pg_stat_database

SELECT * FROM pgperf.snapshot LIMIT 5;
SELECT * FROM pgperf.snapshot_pg_stat_database LIMIT 5;

■ブロック読み込みの推移をスナップショットから取得する

postgres=# SELECT sid,datname,blks_read FROM pgperf.snapshot_pg_stat_database;
 sid |       datname       | blks_read
-----+---------------------+-----------
   2 | template1           |         0
   2 | template0           |         0
   2 | postgres            |     58122
   2 | pgbench             |         0
   2 | snaga               |       138
   2 | testdb              |         0
   2 | testdb0             | 172543546
   3 | template1           |         0
   3 | template0           |         0
   3 | postgres            |     74454
   3 | pgbench             |         0
   3 | snaga               |       138
   3 | testdb              |         0
   3 | testdb0             | 172543546
   4 | template1           |         0
   4 | template0           |         0
   4 | postgres            |     90749
   4 | pgbench             |         0
   4 | snaga               |       138
   4 | testdb              |         0
   4 | testdb0             | 172543546
   5 | template1           |         0
   5 | template0           |         0
   5 | postgres            |    107030
(...snip...)

postgres=# SELECT sid,sum(blks_read) FROM pgperf.snapshot_pg_stat_database GROUP BY sid ORDER BY sid;
 sid |    sum
-----+-----------
   2 | 537034870
   3 | 537051202
   4 | 537067497
   5 | 537083778
   6 | 537100021
(...snip...)

■ウィンドウ関数で差分を取得する

SELECT sid,
       sum(blks_read),
       lag(sum(blks_read)) OVER (ORDER BY sid)
  FROM pgperf.snapshot_pg_stat_database
 GROUP BY sid
 ORDER BY sid;

postgres=# SELECT sid, sum(blks_read), lag(sum(blks_read)) OVER (ORDER BY sid)
  FROM pgperf.snapshot_pg_stat_database GROUP BY sid ORDER BY sid;
 sid  |    sum    |    lag
------+-----------+-----------
    2 | 537034870 |
    3 | 537051202 | 537034870
    4 | 537067497 | 537051202
    5 | 537083778 | 537067497
    6 | 537100021 | 537083778
    7 | 537116221 | 537100021
(...snip...)

SELECT sid,
       sum(blks_read) - lag(sum(blks_read)) OVER (ORDER BY sid) as "blks_read"
  FROM pgperf.snapshot_pg_stat_database
 GROUP BY sid
 ORDER BY sid;

postgres=# SELECT sid,
       sum(blks_read) - lag(sum(blks_read)) OVER (ORDER BY sid) as "blks_read"
  FROM pgperf.snapshot_pg_stat_database
 GROUP BY sid ORDER BY sid;
 sid  | blks_read
------+-----------
    2 |
    3 |     16332
    4 |     16295
    5 |     16281
    6 |     16243
    7 |     16200
(...snip...)

■秒単位の数値に変換する

postgres=# SELECT pgperf.get_interval(2, 3);
 get_interval
--------------
          600
(1 row)

postgres=#

SELECT s.ts,
       ( sum(blks_read) - lag(sum(blks_read)) OVER (ORDER BY s.sid) ) /
         pgperf.get_interval(lag(s.sid) OVER (ORDER BY s.sid), s.sid) as "blks_read/sec"
  FROM pgperf.snapshot_pg_stat_database d, pgperf.snapshot s
 WHERE d.sid = s.sid
 GROUP BY s.sid
 ORDER BY s.ts;

postgres=# SELECT s.ts,
       ( sum(blks_read) - lag(sum(blks_read)) OVER (ORDER BY s.sid) ) /
         pgperf.get_interval(lag(s.sid) OVER (ORDER BY s.sid), s.sid) as "blks_read/sec"
  FROM pgperf.snapshot_pg_stat_database d, pgperf.snapshot s
 WHERE d.sid = s.sid
 GROUP BY s.sid
 ORDER BY s.ts;
             ts             |     blks_read/sec
----------------------------+-----------------------
 2012-11-21 18:20:01.238885 |
 2012-11-21 18:30:01.464424 |   27.2200000000000000
 2012-11-21 18:40:01.685642 |   27.1583333333333333
 2012-11-21 19:21:07.584854 |    6.6021897810218978
 2012-11-21 19:30:01.215188 |   30.4176029962546816
 2012-11-21 19:40:01.442753 |   27.0000000000000000
 2012-11-21 19:50:01.665997 |   26.9500000000000000
(...snip...)

■まとめ

■R、および関連パッケージのインストール

> install.packages("DBI")
> install.packages("RPostgreSQL")

■ブロック書き込み量を取得するビューを作る

CREATE VIEW pgperf.rpt_blkwrite AS
  SELECT date_trunc('second', s.ts) AS "timestamp",
         round( (( buffers_checkpoint - lag(buffers_checkpoint) OVER (ORDER BY s.ts) )::float /
           pgperf.get_interval(lag(s.sid) OVER (ORDER BY s.ts), s.sid))::numeric, 2) AS "blks_cp/sec",
         round( (( buffers_clean - lag(buffers_clean) OVER (ORDER BY s.ts) )::float /
           pgperf.get_interval(lag(s.sid) OVER (ORDER BY s.ts), s.sid))::numeric, 2) AS "blks_bg/sec",
         round( (( buffers_backend - lag(buffers_backend) OVER (ORDER BY s.ts) )::float /
           pgperf.get_interval(lag(s.sid) OVER (ORDER BY s.ts), s.sid))::numeric, 2) AS "blks_be/sec"
    FROM pgperf.snapshot s,
         pgperf.snapshot_pg_stat_bgwriter b
   WHERE s.sid=b.sid
     AND s.ts BETWEEN date_trunc('day', now()) - '32 days'::interval
                  AND date_trunc('day', now()) + '1 day'::interval
 ORDER BY s.ts;

postgres=# SELECT * FROM pgperf.rpt_blkwrite;
      timestamp      | blks_cp/sec | blks_bg/sec | blks_be/sec
---------------------+-------------+-------------+-------------
 2012-12-09 15:45:48 |             |             |
 2012-12-09 15:50:01 |        0.00 |        0.00 |        0.05
 2012-12-09 16:00:01 |        0.00 |        0.00 |        0.00
 2012-12-09 16:10:01 |        0.00 |        0.00 |        0.01
 2012-12-09 16:20:01 |        0.00 |        0.00 |        0.00
(5 rows)

postgres=#

■データベースへ接続してクエリを実行する

> library("RPostgreSQL")
要求されたパッケージ DBI をロード中です 
>

> con <- dbConnect(PostgreSQL(), host="10.0.2.12", user= "postgres", password="postgres", dbname="postgres")
> con
<PostgreSQLConnection:(9088,3)>
>

> rs <- dbSendQuery(con, "SELECT * FROM pgperf.rpt_blkwrite")

■結果を取得して時系列のグラフを描画する

> out <- fetch(rs)
> out
            timestamp blks_cp/sec blks_bg/sec blks_be/sec
1 2012-12-09 15:50:01          NA          NA          NA
2 2012-12-09 16:00:01        0.00           0        0.00
3 2012-12-09 16:10:01        0.00           0        0.01
4 2012-12-09 16:20:01        0.00           0        0.00
5 2012-12-09 16:30:01        0.00           0        0.42
6 2012-12-09 16:40:01       51.36           0        0.00
(...snip...)
>

> ts.plot (out[2:4], gpars=list( ylab='blocks/sec', lty=c(1,1,1), col=c('green', 'red', 'blue')))

> legend('topright', c('Checkpoint','Bgwriter','Backend'), lty=c(1,1,1), col=c('green', 'red', 'blue'))

■まとめ

■参考文献

■「GrowthForecast」とは何か

■GrowthForecastのインストール

[snaga@devsv02 gf]$ wget --no-check-certificate https://github.com/downloads/kazeburo/GrowthForecast/GrowthForecast-0.32.tar.gz
--2012-11-26 23:41:44--  https://github.com/downloads/kazeburo/GrowthForecast/GrowthForecast-0.32.tar.gz
Resolving github.com... 207.97.227.239
(...snip...)
Saving to: `GrowthForecast-0.32.tar.gz'

100%[=============================================>] 91,375      84.5K/s   in 1.1s

2012-11-26 23:41:47 (84.5 KB/s) - `GrowthForecast-0.32.tar.gz' saved [91375/91375]

[snaga@devsv02 gf]$ ls
GrowthForecast-0.32.tar.gz
[snaga@devsv02 gf]$ su
Password: 
[root@devsv02 gf]# cpanm -n GrowthForecast-0.32.tar.gz
Building GrowthForecast-0.32 ... OK
Successfully installed GrowthForecast-0.32
111 distributions installed
[root@devsv02 gf]#

■GrowthForecastサーバの起動

[snaga@devsv02 ~]$ /usr/bin/growthforecast.pl --data-dir /tmp/gfdata
23:15:47 1.1        | 2012-11-27T23:15:47 [INFO] [short] first updater start in Tue Nov 27 23:16:00 2012 at /usr/lib/perl5/site_perl/5.8.8/GrowthForecast/Worker.pm line 55
23:15:47 2.1        | 2012-11-27T23:15:47 [INFO] [update] first updater start in Tue Nov 27 23:20:00 2012 at /usr/lib/perl5/site_perl/5.8.8/GrowthForecast/Worker.pm line 55

■GrowthForecastへのデータの登録

http://ホスト名:ポート番号/api/サービス名/セクション名/グラフ名

http://10.0.2.12:5125/api/postgres/database/session

[snaga@devsv02 gf]$ psql -c 'select count(*) from pg_stat_activity' postgres     count
-------
     1
(1 row)

[snaga@devsv02 gf]$ psql -A -t -c 'select count(*) from pg_stat_activity' postgres
1
[snaga@devsv02 gf]$ curl -F number=`psql -A -t -c 'select count(*) from pg_stat_activity' postgres` http://10.0.2.12:5125/api/postgres/database/session
{"error":0,"data":{"number":1,"llimit":-1000000000,"mode":"gauge","stype":"AREA","adjustval":"1","meta":"","service_name":"postgres","gmode":"gauge","color":"#3399cc","created_at":"2012/11/29 19:00:14","section_name":"database","ulimit":1000000000,"id":2,"graph_name":"session","description":"","sulimit":100000,"unit":"","sort":0,"updated_at":"2012/11/29 19:00:14","adjust":"*","type":"AREA","sllimit":-100000,"md5":"c81e728d9d4c2f636f067f89cc14862c"}}
[snaga@devsv02 gf]$

■バッファの書き出しをグラフ化する

SELECT buffers_checkpoint FROM pg_stat_bgwriter;
SELECT buffers_clean FROM pg_stat_bgwriter;
SELECT buffers_backend FROM pg_stat_bgwriter;

• 差分を計算する
• 複数のグラフを積み上げる

curl -F number=`psql -A -t \
  -c 'SELECT buffers_checkpoint FROM pg_stat_bgwriter' postgres` \
  http://10.0.2.12:5125/api/postgres/database/buffers_checkpoint

curl -F number=`psql -A -t \
  -c 'SELECT buffers_clean FROM pg_stat_bgwriter' postgres` \
  http://10.0.2.12:5125/api/postgres/database/buffers_clean

curl -F number=`psql -A -t \
  -c 'SELECT buffers_backend FROM pg_stat_bgwriter' postgres` \
  http://10.0.2.12:5125/api/postgres/database/buffers_backend

■まとめ

■HinemosとPostgreSQL性能監視

• Hinemosのバージョンが4.0以降であること
• PostgreSQLサーバでHinemosエージェントが動作していること
• PostgreSQLサーバがUnix系サーバであり、Perlが動作すること

■インストール

[snaga@devsv02 hinemos]$ wget --no-check-certificate https://github.com/downloads/uptimejp/hinemos-postgres-addon/hinemos-postgres-addon-r20121114.zip
--2012-12-04 01:44:39--  https://github.com/downloads/uptimejp/hinemos-postgres-addon/hinemos-postgres-addon-r20121114.zip
(...snip...)
Length: 13016 (13K) [application/zip]
Saving to: `hinemos-postgres-addon-r20121114.zip'

100%[======================================>] 13,016      30.1K/s   in 0.4s

2012-12-04 01:44:49 (30.1 KB/s) - `hinemos-postgres-addon-r20121114.zip' saved [13016/13016]

[snaga@devsv02 hinemos]$ 

[snaga@devsv02 hinemos]$ unzip hinemos-postgres-addon-r20121114.zip
Archive:  hinemos-postgres-addon-r20121114.zip
   creating: hinemos-postgres-addon-r20121114/
  inflating: hinemos-postgres-addon-r20121114/test_mon_pgsql.sh
  inflating: hinemos-postgres-addon-r20121114/LICENSE
  inflating: hinemos-postgres-addon-r20121114/Makefile
  inflating: hinemos-postgres-addon-r20121114/mon_pgsql
[snaga@devsv02 hinemos]$ cd hinemos-postgres-addon-r20121114
[snaga@devsv02 hinemos-postgres-addon-r20121114]$ ls -l
total 48
-rw-rw-r-- 1 snaga snaga 18092 Nov 14 13:12 LICENSE
-rw-rw-r-- 1 snaga snaga   682 Nov 14 13:12 Makefile
-rwxrwxr-x 1 snaga snaga 19695 Nov 14 13:12 mon_pgsql*
-rwxrwxr-x 1 snaga snaga   441 Nov 14 13:12 test_mon_pgsql.sh*
[snaga@devsv02 hinemos-postgres-addon-r20121114]$ su
Password:
[root@devsv02 hinemos-postgres-addon-r20121114]# make install
install -m 755 mon_pgsql /opt/hinemos_agent/bin
[root@devsv02 hinemos-postgres-addon-r20121114]# 

[postgres@devsv02 ~]$ /opt/hinemos_agent/bin/mon_pgsql

Usage: /opt/hinemos_agent/bin/mon_pgsql [<option>...] <search_key>

Options:
    -h <host>     Host name to connect the database.
    -p <port>     Port number to connect the database.
    -U <user>     User name used to connect the database.
    -d <dbname>   Database name to be connected.
    -v            Verbose output.

Search keys:
    session
    cache_hit
    tuple_wrtn
    tuple_read
    blks_read
    blks_wrtn
    txn
    xlog_wrtn
    dbsize
    locks

[postgres@devsv02 ~]$ 

• session：セッション数
• cache_hit：キャッシュヒット率
• tuple_wrtn：タプル更新（INSERT/UPDATE/DELETE）
• tuple_read：タプル読み込み（RETURN/FETCH）
• blks_read：ブロック読み込み
• blks_wrtn：ブロック書き込み
• txn：トランザクション数
• xlog_wrtn：WAL書き込み量
• dbsize：オブジェクトサイズ
• locks：ロック

[postgres@devsv02 ~]$ /opt/hinemos_agent/bin/mon_pgsql -h localhost -p 5432 -U postgres -d postgres session
active,1
idle,1
wait,0
[postgres@devsv02 ~]$

■Hinemosでのリポジトリへのノード登録

■Hinemosでのカスタム監視の登録

• 監視項目ID：監視の設定を識別する任意のIDを指定します。
• スコープ：監視対象（スコープ）を指定します。
• 間隔：監視の間隔を指定します。
• 実効ユーザ：mon_pgsqlスクリプトを実行するOSユーザを指定します。
• コマンド：mon_psgqlスクリプトを実行するコマンドラインを指定します。
• 監視／判定（情報、警告、危険）：監視で設定する閾値を入力します。
• 監視／アプリケーション：アプリケーション名を記入します。
• 収集／収集値表示名：収集・蓄積するデータの名称を記入します。
• 収集／収集値単位：収集・蓄積するデータの単位を記入します。

■Hinemosでの性能情報の表示

■まとめ

• 【リリース情報】セキュリティリリース（9.2.4、9.1.9、9.0.13、8.4.17）が出ています
  http://pgsqldeepdive.blogspot.jp/2013/04/securityrelease.html

■Amazon EC2でインスタンスを立ち上げる

• Instance Type : T1 Micro
• Availability Zone : No Preference
• Key Pair : Create a new Key Pair （または既存のものがあればそれを選ぶ）

• Security Group : Create a new Security Group
• Group Name : test-pgsql
• Group Description : test-pgsql
• Create a new rule : SSH
• Source : 0.0.0.0/0 （または制限できるのであれば指定する）
• Create a new rule : Custom TCP rule
• Port range : 16543
• Source : 0.0.0.0/0 （または制限できるのであれば指定する）

login as: ec2-user
Authenticating with public key "imported-openssh-key"

       __|  __|_  )
       _|  (     /   Amazon Linux AMI
      ___|\___|___|

https://aws.amazon.com/amazon-linux-ami/2012.09-release-notes/
There are 4 security update(s) out of 33 total update(s) available
Run "sudo yum update" to apply all updates.
[ec2-user@ip-10-152-167-242 ~]$ 

[ec2-user@ip-10-152-167-242 ~]$ sudo su
[root@ip-10-152-167-242 ec2-user]# yum update
Loaded plugins: priorities, security, update-motd, upgrade-helper
Setting up Update Process
Resolving Dependencies
(...snip...)
  ruby-libs.i686 0:1.8.7.371-1.20.amzn1     tzdata.noarch 0:2012f-1.15.amzn1
  tzdata-java.noarch 0:2012f-1.15.amzn1     yum.noarch 0:3.2.29-30.24.amzn1

Complete!
[root@ip-10-152-167-242 ec2-user]#

[root@ip-10-152-167-242 ~]# yum install libxslt.i686 uuid.i686
Loaded plugins: priorities, security, update-motd, upgrade-helper
Setting up Install Process
Resolving Dependencies
--> Running transaction check
---> Package libxslt.i686 0:1.1.26-2.7.amzn1 will be installed
---> Package uuid.i686 0:1.6.2-11.16.amzn1 will be installed
(...snip...)
Installed:
  libxslt.i686 0:1.1.26-2.7.amzn1
  uuid.i686 0:1.6.2-11.16.amzn1

Complete!
[root@ip-10-152-167-242 ~]# 

■PostgreSQLをインストールする

[root@ip-10-152-167-242 ~]# sh dl.sh
--2012-11-24 09:15:20--  http://yum.postgresql.org/9.2/redhat/rhel-6.3-i386/postgresql92-9.2.1-1PGDG.rhel6.i686.rpm
Resolving yum.postgresql.org... 98.129.198.114
Connecting to yum.postgresql.org|98.129.198.114|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 959964 (937K) [application/x-redhat-package-manager]
Saving to: “postgresql92-9.2.1-1PGDG.rhel6.i686.rpm”

100%[======================================>] 959,964      634K/s   in 1.5s

2012-11-24 09:15:22 (634 KB/s) - “postgresql92-9.2.1-1PGDG.rhel6.i686.rpm” saved [959964/959964]

(...snip...)

2012-11-24 09:15:48 (702 KB/s) - “postgresql92-test-9.2.1-1PGDG.rhel6.i686.rpm” saved [1275524/1275524]

[root@ip-10-152-167-242 ~]# ls *.rpm
postgresql92-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-contrib-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-debuginfo-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-devel-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-docs-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-libs-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-plperl-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-plpython-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-pltcl-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-server-9.2.1-1PGDG.rhel6.i686.rpm
postgresql92-test-9.2.1-1PGDG.rhel6.i686.rpm
[root@ip-10-152-167-242 ~]#

• postgresql92-9.2.1-1PGDG.rhel6.i686.rpm
• postgresql92-contrib-9.2.1-1PGDG.rhel6.i686.rpm
• postgresql92-devel-9.2.1-1PGDG.rhel6.i686.rpm
• postgresql92-libs-9.2.1-1PGDG.rhel6.i686.rpm
• postgresql92-server-9.2.1-1PGDG.rhel6.i686.rpm

[root@ip-10-152-167-242 ~]# rpm -ivh postgresql92-9.2.1-1PGDG.rhel6.i686.rpm postgresql92-contrib-9.2.1-1PGDG.rhel6.i686.rpm postgresql92-devel-9.2.1-1PGDG.rhel6.i686.rpm postgresql92-libs-9.2.1-1PGDG.rhel6.i686.rpm postgresql92-server-9.2.1-1PGDG.rhel6.i686.rpm
warning: postgresql92-9.2.1-1PGDG.rhel6.i686.rpm: Header V4 DSA/SHA1 Signature, key ID 442df0f8: NOKEY
Preparing...                ########################################### [100%]
   1:postgresql92-libs      ########################################### [ 20%]
   2:postgresql92           ########################################### [ 40%]
   3:postgresql92-contrib   ########################################### [ 60%]
   4:postgresql92-devel     ########################################### [ 80%]
   5:postgresql92-server    ########################################### [100%]
[root@ip-10-152-167-242 ~]#

■データベースクラスタを初期化する

[root@ip-10-152-167-242 ~]# su - postgres
-bash-4.1$ pwd
/var/lib/pgsql
-bash-4.1$ ls -F
9.2/
-bash-4.1$ /usr/pgsql-9.2/bin/initdb -D /var/lib/pgsql/9.2/data --no-locale -E UTF-8
The files belonging to this database system will be owned by user "postgres".
This user must also own the server process.

The database cluster will be initialized with locale "C".
The default text search configuration will be set to "english".

fixing permissions on existing directory /var/lib/pgsql/9.2/data ... ok
(...snip...)

Success. You can now start the database server using:

    /usr/pgsql-9.2/bin/postgres -D /var/lib/pgsql/9.2/data
or
    /usr/pgsql-9.2/bin/pg_ctl -D /var/lib/pgsql/9.2/data -l logfile start

-bash-4.1$

■サーバの鍵と証明書を作成する

-bash-4.1$ openssl genrsa -out server.key 1024
Generating RSA private key, 1024 bit long modulus
.........................................++++++
....++++++
e is 65537 (0x10001)
-bash-4.1$ openssl req -new -key server.key -x509 -days 365 -out server.crt
(...snip...)
Country Name (2 letter code) [XX]:JP
State or Province Name (full name) []:Tokyo
Locality Name (eg, city) [Default City]:Minato-ku
Organization Name (eg, company) [Default Company Ltd]:Uptime Technologies, LLC
Organizational Unit Name (eg, section) []:
Common Name (eg, your name or your server's hostname) []:ec2-54-248-9-93.ap-northeast-1.compute.amazonaws.com
Email Address []:nobody@uptime.jp
-bash-4.1$ ls -F
9.2/  server.crt  server.key
-bash-4.1$

-bash-4.1$ chmod 600 server.*
-bash-4.1$ mv server.* 9.2/data/
-bash-4.1$ cd 9.2/data/
-bash-4.1$ ls -F
base/        pg_ident.conf  pg_serial/     pg_tblspc/    postgresql.conf
global/      pg_log/        pg_snapshots/  pg_twophase/  postmaster.opts
pg_clog/     pg_multixact/  pg_stat_tmp/   PG_VERSION    server.crt
pg_hba.conf  pg_notify/     pg_subtrans/   pg_xlog/      server.key
-bash-4.1$

■PostgreSQLにSSLの設定行い、サービスを起動する

listen_addresses = '*'
port = 16543
ssl = on
ssl_cert_file = 'server.crt'
ssl_key_file = 'server.key'

hostssl all             all             0.0.0.0/0               password

PGPORT=16543

[root@ip-10-152-167-242 ~]# /sbin/chkconfig postgresql-9.2 on
[root@ip-10-152-167-242 ~]# /sbin/chkconfig --list postgresql-9.2
postgresql-9.2  0:off   1:off   2:on    3:on    4:on    5:on    6:off
[root@ip-10-152-167-242 ~]# /etc/rc.d/init.d/postgresql-9.2 start
Starting postgresql-9.2 service:                           [  OK  ]
[root@ip-10-152-167-242 ~]#

■ユーザを作成し、パスワードを設定する

[root@ip-10-152-167-242 ~]# su - postgres
-bash-4.1$ createuser -p 16543 --createdb --encrypted --pwprompt snaga
Enter password for new role:
Enter it again:
-bash-4.1$

■pgAdminIIIから接続する

■まとめ

■検証構成

• Amazon EC2 Linux AMI (32bit)
• PostgreSQL 9.2
• IPアドレス：10.156.93.165

• Amazon EC2 Linux AMI (32bit)
• PostgreSQL 9.2
• IPアドレス：10.146.98.58
• 受信したWALを蓄積するディレクトリ：/var/lib/pgsql/9.2/receivexlog/
• ベースバックアップを取得するディレクトリ：/var/lib/pgsql/9.2/receivebase/

■pg_receivexlogのための設定項目

wal_level = archive
archive_command = 'cat /dev/null'
max_wal_senders = 3
wal_keep_segments = 8

hostssl replication     postgres        10.146.98.58/32         trust

■リアルタイムバックアップを開始する

-bash-4.1$ /usr/pgsql-9.2/bin/pg_receivexlog -h 10.156.93.165 -p 6453 -U postgres -D /var/lib/pgsql/9.2/receivexlog -v
pg_receivexlog: starting log streaming at 0/A000000 (timeline 1)

[root@ip-10-146-98-58 ~]# ls -al /var/lib/pgsql/9.2/receivexlog/
total 16392
drwxr-xr-x 2 postgres postgres     4096 Dec  6 14:40 .
drwx------ 5 postgres postgres     4096 Dec  6 14:33 ..
-rw------- 1 postgres postgres 16777216 Dec  6 14:40 00000001000000000000000A.partial
[root@ip-10-146-98-58 ~]# 

-bash-4.1$ /usr/pgsql-9.2/bin/pgbench -p 6453 -s 10 -i testdb

[root@ip-10-146-98-58 ~]# ls -al /var/lib/pgsql/9.2/receivexlog/
total 131080
drwxr-xr-x 2 postgres postgres     4096 Dec  6 14:43 .
drwx------ 5 postgres postgres     4096 Dec  6 14:33 ..
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000A
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000B
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000C
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000D
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000E
-rw------- 1 postgres postgres 16777216 Dec  6 14:42 00000001000000000000000F
-rw------- 1 postgres postgres 16777216 Dec  6 14:43 000000010000000000000010
-rw------- 1 postgres postgres 16777216 Dec  6 14:43 000000010000000000000011.partial
[root@ip-10-146-98-58 ~]#

■バックアップセットを取ってみる

bash-4.1$ /usr/pgsql-9.2/bin/pg_receivexlog -h 10.156.93.165 -p 6453 -U postgres -D /var/lib/pgsql/9.2/receivexlog -v
pg_receivexlog: starting log streaming at 0/19000000 (timeline 1)

-bash-4.1$ /usr/pgsql-9.2/bin/pg_basebackup  -h 10.156.93.165 -p 6453 -U postgres -D /var/lib/pgsql/9.2/receivebase -v
NOTICE:  pg_stop_backup complete, all required WAL segments have been archived
pg_basebackup: base backup completed
-bash-4.1$

[snaga@db01 ~]$ psql -p 6453 testdb
psql (9.2.1)
Type "help" for help.

testdb=> \d
               List of relations
 Schema |        Name        | Type  |  Owner
--------+--------------------+-------+----------
 public | pg_stat_statements | view  | postgres
 public | pgbench_accounts   | table | postgres
 public | pgbench_branches   | table | postgres
 public | pgbench_history    | table | postgres
 public | pgbench_tellers    | table | postgres
 public | t1                 | table | postgres
 public | t2                 | table | snaga
(7 rows)

testdb=> create table t3 ( uid integer primary key, uname text not null );
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "t3_pkey" for table "t3"
CREATE TABLE
testdb=> insert into t3 values ( 101, 'Jung, Nicole' );
INSERT 0 1
testdb=>

bash-4.1$ /usr/pgsql-9.2/bin/pg_receivexlog -h 10.156.93.165 -p 6453 -U postgres -D /var/lib/pgsql/9.2/receivexlog -v
pg_receivexlog: starting log streaming at 0/19000000 (timeline 1)
pg_receivexlog: finished segment at 0/1A000000 (timeline 1)
^Cpg_receivexlog: received interrupt signal, exiting.
pg_receivexlog: not renaming "00000001000000000000001A", segment is not complete
bash-4.1$

■アーカイブログリカバリを実施

-bash-4.1$ pwd
/var/lib/pgsql
-bash-4.1$ ls -F
9.2/
-bash-4.1$ cd 9.2/
-bash-4.1$ ls -F
backups/  data/  receivebase/  receivexlog/
-bash-4.1$ ls data/
-bash-4.1$ cp -r receivebase/* data/
-bash-4.1$ ls -F data
backup_label  pg_ident.conf  pg_snapshots/  PG_VERSION            server.key
base/         pg_log/        pg_stat_tmp/   pg_xlog/
global/       pg_multixact/  pg_subtrans/   postgresql.conf
pg_clog/      pg_notify/     pg_tblspc/     postgresql.conf.orig
pg_hba.conf   pg_serial/     pg_twophase/   server.crt

-bash-4.1$ cp receivexlog/0000000100000000000000* data/pg_xlog/
-bash-4.1$ cd data/pg_xlog/
-bash-4.1$ ls -F
00000001000000000000000A  000000010000000000000013
00000001000000000000000B  000000010000000000000014
00000001000000000000000C  000000010000000000000015
00000001000000000000000D  000000010000000000000016
00000001000000000000000E  000000010000000000000017
00000001000000000000000F  000000010000000000000018
000000010000000000000010  000000010000000000000019
000000010000000000000011  00000001000000000000001A.partial
000000010000000000000012
-bash-4.1$ mv 00000001000000000000001A.partial 00000001000000000000001A
-bash-4.1$ pwd
/var/lib/pgsql/9.2/data/pg_xlog
-bash-4.1$

[root@ip-10-146-98-58 ~]# /etc/rc.d/init.d/postgresql-9.2 start
Starting postgresql-9.2 service:                           [  OK  ]
[root@ip-10-146-98-58 ~]#

[root@ip-10-146-98-58 ~]# su - postgres
-bash-4.1$ psql testdb
psql (9.2.1)
Type "help" for help.

testdb=# \d
               List of relations
 Schema |        Name        | Type  |  Owner
--------+--------------------+-------+----------
 public | pg_stat_statements | view  | postgres
 public | pgbench_accounts   | table | postgres
 public | pgbench_branches   | table | postgres
 public | pgbench_history    | table | postgres
 public | pgbench_tellers    | table | postgres
 public | t1                 | table | postgres
 public | t2                 | table | snaga
 public | t3                 | table | snaga
(8 rows)

testdb=# select * from t3;
 uid |    uname
-----+--------------
 101 | Jung, Nicole
(1 row)

testdb=#

[2012-12-07 00:10:14 JST] 4197: LOG:  database system was interrupted; last known up at 2012-12-07 00:00:16 JST
[2012-12-07 00:10:14 JST] 4197: LOG:  creating missing WAL directory "pg_xlog/archive_status"
[2012-12-07 00:10:14 JST] 4197: LOG:  database system was not properly shut down; automatic recovery in progress
[2012-12-07 00:10:14 JST] 4197: LOG:  redo starts at 0/19000074
[2012-12-07 00:10:14 JST] 4197: LOG:  record with zero length at 0/1A023DEC
[2012-12-07 00:10:14 JST] 4197: LOG:  redo done at 0/1A023DC4
[2012-12-07 00:10:14 JST] 4197: LOG:  last completed transaction was at log time 2012-12-07 00:04:39.920898+09
[2012-12-07 00:10:14 JST] 4194: LOG:  database system is ready to accept connections

■まとめ

■tablelogモジュール

■tablelogモジュールのインストール

[snaga@devsv02 pgsql]$ tar zxf table_log-0.4.4.tar.gz
[snaga@devsv02 pgsql]$ cd table_log-0.4.4
[snaga@devsv02 table_log-0.4.4]$ ls
Makefile          table_log.c  table_log_init.sql  tests
README.table_log  table_log.h  table_log.sql.in
[snaga@devsv02 table_log-0.4.4]$ patch -p1 < ../table_log-0_4_4_snaga.diff
patching file Makefile
patching file table_log.sql.in
patching file table_log_uninstall.sql.in
[snaga@devsv02 table_log-0.4.4]$ env USE_PGXS=1 PATH=/usr/pgsql-9.1/bin:$PATH make
sed 's,MODULE_PATHNAME,$libdir/table_log,g' table_log.sql.in >table_log.sql
sed 's,MODULE_PATHNAME,$libdir/table_log_uninstall,g' table_log_uninstall.sql.in >table_log_uninstall.sql
gcc -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -I/usr/include/et -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Wendif-labels -Wformat-security -fno-strict-aliasing -fwrapv -fpic -I. -I. -I/usr/pgsql-9.1/include/server -I/usr/pgsql-9.1/include/internal -I/usr/include/et -D_GNU_SOURCE -I/usr/include/libxml2  -I/usr/include  -c -o table_log.o table_log.c
gcc -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -m64 -mtune=generic -I/usr/include/et -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Wendif-labels -Wformat-security -fno-strict-aliasing -fwrapv -fpic -L/usr/pgsql-9.1/lib -L/usr/lib64  -shared -o table_log.so table_log.o
rm table_log.o
[snaga@devsv02 table_log-0.4.4]$ su
Password:
[root@devsv02 table_log-0.4.4]# env USE_PGXS=1 PATH=/usr/pgsql-9.1/bin:$PATH make install
/bin/mkdir -p '/usr/pgsql-9.1/share/contrib'
/bin/mkdir -p '/usr/pgsql-9.1/lib'
/bin/mkdir -p '/usr/share/doc/pgsql/contrib'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 644  table_log.sql table_log_uninstall.sql '/usr/pgsql-9.1/share/contrib/'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 755  table_log.so '/usr/pgsql-9.1/lib/'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 644 ./README.table_log '/usr/share/doc/pgsql/contrib/'
[root@devsv02 table_log-0.4.4]# exit
exit
[snaga@devsv02 table_log-0.4.4]$

[snaga@devsv02 table_log-0.4.4]$ psql -f /usr/pgsql-9.1/share/contrib/table_log.sql testdb2
SET
CREATE FUNCTION
CREATE FUNCTION
(...snip...)
CREATE FUNCTION
CREATE FUNCTION
[snaga@devsv02 table_log-0.4.4]$

■ターゲットとなるテーブルにロギングを設定する

testdb2=# CREATE table t1 ( uid integer primary key, uname text not null );
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "t1_pkey" for table "t1"
CREATE TABLE
testdb2=#

testdb2=# SELECT table_log_init(3, 't1');
 table_log_init
----------------

(1 row)

testdb2=# \d+
                    List of relations
 Schema |  Name  | Type  | Owner |  Size   | Description
--------+--------+-------+-------+---------+-------------
 public | t1     | table | snaga | 0 bytes |
 public | t1_log | table | snaga | 0 bytes |
(2 rows)

testdb2=#

■テーブルを更新してログを取得する

testdb2=# SELECT * FROM t1;
 uid | uname
-----+-------
(0 rows)

testdb2=# INSERT INTO t1 VALUES ( 101, 'satoshi nagayasu' );
INSERT 0 1
testdb2=# SELECT * FROM t1;
 uid |      uname
-----+------------------
 101 | satoshi nagayasu
(1 row)

testdb2=# SELECT * FROM t1_log;
 uid |      uname       | trigger_mode | trigger_tuple |        trigger_changed
-----+------------------+--------------+---------------+-------------------------------
 101 | satoshi nagayasu | INSERT       | new           | 2012-11-26 13:34:01.218549+09
(1 row)

testdb2=# 

testdb2=# UPDATE t1 SET uname = upper(uname);
UPDATE 1
testdb2=# SELECT * FROM t1;
 uid |      uname
-----+------------------
 101 | SATOSHI NAGAYASU
(1 row)

testdb2=# SELECT * FROM t1_log;
 uid |      uname       | trigger_mode | trigger_tuple |        trigger_changed
-----+------------------+--------------+---------------+-------------------------------
 101 | satoshi nagayasu | INSERT       | new           | 2012-11-26 13:34:01.218549+09
 101 | satoshi nagayasu | UPDATE       | old           | 2012-11-26 13:34:38.38462+09
 101 | SATOSHI NAGAYASU | UPDATE       | new           | 2012-11-26 13:34:38.38462+09
(3 rows)

testdb2=#

■ロギングの停止

testdb2=# SELECT disable_table_log('t1');
 disable_table_log
-------------------

(1 row)

testdb2=# \d t1+
      Table "public.t1"
 Column |  Type   | Modifiers
--------+---------+-----------
 uid    | integer | not null
 uname  | text    | not null
Indexes:
    "t1_pkey" PRIMARY KEY, btree (uid)

testdb2=#

■ログレベルの違い

testdb2=# \x
Expanded display is on.
testdb2=# SELECT * FROM t1_log;
-[ RECORD 1 ]---+------------------------------
uid             | 101
uname           | satoshi nagayasu
trigger_mode    | INSERT
trigger_tuple   | new
trigger_changed | 2012-11-26 13:48:24.141572+09

testdb2=# SELECT * FROM t2_log;
-[ RECORD 1 ]---+------------------------------
uid             | 101
uname           | satoshi nagayasu
trigger_mode    | INSERT
trigger_tuple   | new
trigger_changed | 2012-11-26 13:48:24.142149+09
trigger_id      | 1

testdb2=# SELECT * FROM t3_log;
-[ RECORD 1 ]---+------------------------------
uid             | 101
uname           | satoshi nagayasu
trigger_mode    | INSERT
trigger_tuple   | new
trigger_changed | 2012-11-26 13:48:24.142683+09
trigger_id      | 1
trigger_user    | snaga

testdb2=#

■まとめ

■「追記型のストレージアーキテクチャ」とは

■pageinspectモジュールのインストール

■ブロック内部における新規レコードの状態

testdb=# CREATE TABLE t1 ( uid INTEGER PRIMARY KEY, uname TEXT NOT NULL );
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "t1_pkey" for table "t1"
CREATE TABLE
testdb=#
testdb=# INSERT INTO t1 VALUES ( 101, 'insert 1' );
INSERT 0 1
testdb=# select * from t1;
 uid |  uname
-----+----------
 101 | insert 1
(1 row)

testdb=#

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   8152 |        1 |     37 |   1859 |      0
(1 row)

testdb=#

■レコードに対する更新処理

testdb=# UPDATE t1 SET uname = 'update 1' WHERE uid = 101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   8152 |        1 |     37 |   1859 |   1860
  2 |   8112 |        1 |     37 |   1860 |      0
(2 rows)

testdb=#

testdb=# UPDATE t1 SET uname = 'update 2' WHERE uid = 101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   8152 |        1 |     37 |   1859 |   1860
  2 |   8112 |        1 |     37 |   1860 |   1861
  3 |   8072 |        1 |     37 |   1861 |      0
(3 rows)

testdb=#

■不要領域の解放（VACUUM処理）

testdb=# VACUUM t1;
VACUUM
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |      3 |        2 |      0 |        |
  2 |      0 |        0 |      0 |        |
  3 |   8152 |        1 |     37 |   1861 |      0
(3 rows)

testdb=#

testdb=# UPDATE t1 SET uname = 'update 3' WHERE uid = 101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |      3 |        2 |      0 |        |
  2 |   8112 |        1 |     37 |   1862 |      0
  3 |   8152 |        1 |     37 |   1861 |   1862
(3 rows)

testdb=#

■まとめ

■追記型ストレージにおけるファイルの肥大化とその抑制

■「Page Pruning」とは

■更新処理とブロック内の未使用領域の減少

testdb=# CREATE TABLE t1 ( uid INTEGER PRIMARY KEY, uname TEXT NOT NULL );
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "t1_pkey" for table "t1"
testdb=# INSERT INTO t1 VALUES ( 101, 'insert 1-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------' );
INSERT 0 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  39586 |      0
(1 row)

testdb=# UPDATE t1 SET uname = 'update 0-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------' WHERE uid=101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  39586 |  39587
  2 |   7568 |        1 |    309 |  39587 |      0
(2 rows)

testdb=# UPDATE t1 SET uname = 'update 1-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------' WHERE uid=101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  39586 |  39587
  2 |   7568 |        1 |    309 |  39587 |  39588
  3 |   7256 |        1 |    309 |  39588 |      0
(3 rows)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
             8192
(1 row)

testdb=#

 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  39586 |  39587
  2 |   7568 |        1 |    309 |  39587 |  39588
  3 |   7256 |        1 |    309 |  39588 |  39589
  4 |   6944 |        1 |    309 |  39589 |  39590
  5 |   6632 |        1 |    309 |  39590 |  39591
  6 |   6320 |        1 |    309 |  39591 |  39592
  7 |   6008 |        1 |    309 |  39592 |  39593
  8 |   5696 |        1 |    309 |  39593 |  39594
  9 |   5384 |        1 |    309 |  39594 |  39595
 10 |   5072 |        1 |    309 |  39595 |  39596
 11 |   4760 |        1 |    309 |  39596 |  39597
 12 |   4448 |        1 |    310 |  39597 |  39598
 13 |   4136 |        1 |    310 |  39598 |  39599
 14 |   3824 |        1 |    310 |  39599 |  39600
 15 |   3512 |        1 |    310 |  39600 |  39601
 16 |   3200 |        1 |    310 |  39601 |  39602
 17 |   2888 |        1 |    310 |  39602 |  39603
 18 |   2576 |        1 |    310 |  39603 |  39604
 19 |   2264 |        1 |    310 |  39604 |  39605
 20 |   1952 |        1 |    310 |  39605 |  39606
 21 |   1640 |        1 |    310 |  39606 |  39607
 22 |   1328 |        1 |    310 |  39607 |  39608
 23 |   1016 |        1 |    310 |  39608 |  39609
 24 |    704 |        1 |    310 |  39609 |      0
(24 rows)

■Page Pruning処理の実施

 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |     24 |        2 |      0 |        |
  2 |   7568 |        1 |    310 |  39610 |      0
  3 |      0 |        0 |      0 |        |
  4 |      0 |        0 |      0 |        |
  5 |      0 |        0 |      0 |        |
  6 |      0 |        0 |      0 |        |
  7 |      0 |        0 |      0 |        |
  8 |      0 |        0 |      0 |        |
  9 |      0 |        0 |      0 |        |
 10 |      0 |        0 |      0 |        |
 11 |      0 |        0 |      0 |        |
 12 |      0 |        0 |      0 |        |
 13 |      0 |        0 |      0 |        |
 14 |      0 |        0 |      0 |        |
 15 |      0 |        0 |      0 |        |
 16 |      0 |        0 |      0 |        |
 17 |      0 |        0 |      0 |        |
 18 |      0 |        0 |      0 |        |
 19 |      0 |        0 |      0 |        |
 20 |      0 |        0 |      0 |        |
 21 |      0 |        0 |      0 |        |
 22 |      0 |        0 |      0 |        |
 23 |      0 |        0 |      0 |        |
 24 |   7880 |        1 |    310 |  39609 |  39610
(24 rows)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
             8192
(1 row)

testdb=# 

■まとめ

■インデックスの内部構造

testdb=# INSERT INTO t1 VALUES ( 101, 'insert 1' );
INSERT 0 1
testdb=# 
testdb=# SELECT * FROM bt_page_items('t1_pkey', 1);
 itemoffset | ctid  | itemlen | nulls | vars |    data
------------+-------+---------+-------+------+-------------
          1 | (0,1) |      12 | f     | f    | 65 00 00 00
(1 row)

testdb=#

■インデックスとレコードの関連付け

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  43737 |      0
(1 row)

testdb=#

■レコードに対する更新処理（HOT Update）

testdb=# INSERT INTO t1 VALUES ( 101, 'insert 1                                                                                                                                                                                                                                                                             ' );
INSERT 0 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  42731 |      0
(1 row)

testdb=# select * from bt_page_items('t1_pkey', 1);
 itemoffset | ctid  | itemlen | nulls | vars |    data
------------+-------+---------+-------+------+-------------
          1 | (0,1) |      12 | f     | f    | 65 00 00 00
(1 row)

testdb=# 

testdb=# UPDATE t1 SET uname = 'update 0-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------' WHERE uid=101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  42731 |  42732
  2 |   7568 |        1 |    309 |  42732 |      0
(2 rows)

testdb=# select * from bt_page_items('t1_pkey', 1);
 itemoffset | ctid  | itemlen | nulls | vars |    data
------------+-------+---------+-------+------+-------------
          1 | (0,1) |      12 | f     | f    | 65 00 00 00
(1 row)

testdb=#

• インデックスキー（dataカラムの値）からレコードの位置（ctidカラムの値）を取得
• ctidカラムから読みだしたテーブル内の位置のレコード（lp==1）を取得
• 当該レコードは削除済（t_xmax==42732）のため、t_xmaxとt_xminのチェーンを辿ってlp==2のレコードを見つける
• lp==2はまだ削除されていない（t_xmaxが空）レコードなので、これを使う

■Page Pruning処理とインデックス

 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    309 |  42731 |  42732
  2 |   7568 |        1 |    309 |  42732 |  42733
  3 |   7256 |        1 |    309 |  42733 |  42734
  4 |   6944 |        1 |    309 |  42734 |  42735
  5 |   6632 |        1 |    309 |  42735 |  42736
  6 |   6320 |        1 |    309 |  42736 |  42737
  7 |   6008 |        1 |    309 |  42737 |  42738
  8 |   5696 |        1 |    309 |  42738 |  42739
  9 |   5384 |        1 |    309 |  42739 |  42740
 10 |   5072 |        1 |    309 |  42740 |  42741
 11 |   4760 |        1 |    309 |  42741 |  42742
 12 |   4448 |        1 |    310 |  42742 |  42743
 13 |   4136 |        1 |    310 |  42743 |  42744
 14 |   3824 |        1 |    310 |  42744 |  42745
 15 |   3512 |        1 |    310 |  42745 |  42746
 16 |   3200 |        1 |    310 |  42746 |  42747
 17 |   2888 |        1 |    310 |  42747 |  42748
 18 |   2576 |        1 |    310 |  42748 |  42749
 19 |   2264 |        1 |    310 |  42749 |  42750
 20 |   1952 |        1 |    310 |  42750 |  42751
 21 |   1640 |        1 |    310 |  42751 |  42752
 22 |   1328 |        1 |    310 |  42752 |  42753
 23 |   1016 |        1 |    310 |  42753 |  42754
 24 |    704 |        1 |    310 |  42754 |      0
(24 rows)

 itemoffset | ctid  | itemlen | nulls | vars |    data
------------+-------+---------+-------+------+-------------
          1 | (0,1) |      12 | f     | f    | 65 00 00 00
(1 row)

 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |     24 |        2 |      0 |        |
  2 |   7568 |        1 |    310 |  42755 |      0
  3 |      0 |        0 |      0 |        |
  4 |      0 |        0 |      0 |        |
  5 |      0 |        0 |      0 |        |
  6 |      0 |        0 |      0 |        |
  7 |      0 |        0 |      0 |        |
  8 |      0 |        0 |      0 |        |
  9 |      0 |        0 |      0 |        |
 10 |      0 |        0 |      0 |        |
 11 |      0 |        0 |      0 |        |
 12 |      0 |        0 |      0 |        |
 13 |      0 |        0 |      0 |        |
 14 |      0 |        0 |      0 |        |
 15 |      0 |        0 |      0 |        |
 16 |      0 |        0 |      0 |        |
 17 |      0 |        0 |      0 |        |
 18 |      0 |        0 |      0 |        |
 19 |      0 |        0 |      0 |        |
 20 |      0 |        0 |      0 |        |
 21 |      0 |        0 |      0 |        |
 22 |      0 |        0 |      0 |        |
 23 |      0 |        0 |      0 |        |
 24 |   7880 |        1 |    310 |  42754 |  42755
(24 rows)

 itemoffset | ctid  | itemlen | nulls | vars |    data
------------+-------+---------+-------+------+-------------
          1 | (0,1) |      12 | f     | f    | 65 00 00 00
(1 row)

• インデックスの ctid==(0,1) からテーブルのlp==1のレコードを見る
• lp==1がリダイレクトなので、lp_off==24からラインポインタ24番（lp==24）を見る
• lp==24は削除済（t_xmax==42755）なので、チェーンをたどって次の新しいレコード（lp==2）を探す
• lp==2はまだ生きているので、このレコードを使う

■まとめ

■ブロックに空き領域が無い時の更新処理

testdb=# INSERT INTO t1 VALUES ( 125, 'insert 125                                                                                                                                                                                                                                                                             ' );
INSERT 0 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46798 |      0
  2 |   7568 |        1 |    311 |  46799 |      0
  3 |   7256 |        1 |    311 |  46800 |      0
  4 |   6944 |        1 |    311 |  46801 |      0
  5 |   6632 |        1 |    311 |  46802 |      0
  6 |   6320 |        1 |    311 |  46803 |      0
  7 |   6008 |        1 |    311 |  46804 |      0
  8 |   5696 |        1 |    311 |  46805 |      0
  9 |   5384 |        1 |    311 |  46806 |      0
 10 |   5072 |        1 |    311 |  46807 |      0
 11 |   4760 |        1 |    311 |  46808 |      0
 12 |   4448 |        1 |    311 |  46809 |      0
 13 |   4136 |        1 |    311 |  46810 |      0
 14 |   3824 |        1 |    311 |  46811 |      0
 15 |   3512 |        1 |    311 |  46812 |      0
 16 |   3200 |        1 |    311 |  46813 |      0
 17 |   2888 |        1 |    311 |  46814 |      0
 18 |   2576 |        1 |    311 |  46815 |      0
 19 |   2264 |        1 |    311 |  46816 |      0
 20 |   1952 |        1 |    311 |  46817 |      0
 21 |   1640 |        1 |    311 |  46818 |      0
 22 |   1328 |        1 |    311 |  46819 |      0
 23 |   1016 |        1 |    311 |  46820 |      0
 24 |    704 |        1 |    311 |  46821 |      0
 25 |    392 |        1 |    311 |  46822 |      0
(25 rows)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
             8192
(1 row)

testdb=# 

testdb=# UPDATE t1 SET uname = 'update 101                                                                                                                                                                                                                                                                             ' WHERE uid=101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46798 |  46823
  2 |   7568 |        1 |    311 |  46799 |      0
  3 |   7256 |        1 |    311 |  46800 |      0
  4 |   6944 |        1 |    311 |  46801 |      0
  5 |   6632 |        1 |    311 |  46802 |      0
  6 |   6320 |        1 |    311 |  46803 |      0
  7 |   6008 |        1 |    311 |  46804 |      0
  8 |   5696 |        1 |    311 |  46805 |      0
  9 |   5384 |        1 |    311 |  46806 |      0
 10 |   5072 |        1 |    311 |  46807 |      0
 11 |   4760 |        1 |    311 |  46808 |      0
 12 |   4448 |        1 |    311 |  46809 |      0
 13 |   4136 |        1 |    311 |  46810 |      0
 14 |   3824 |        1 |    311 |  46811 |      0
 15 |   3512 |        1 |    311 |  46812 |      0
 16 |   3200 |        1 |    311 |  46813 |      0
 17 |   2888 |        1 |    311 |  46814 |      0
 18 |   2576 |        1 |    311 |  46815 |      0
 19 |   2264 |        1 |    311 |  46816 |      0
 20 |   1952 |        1 |    311 |  46817 |      0
 21 |   1640 |        1 |    311 |  46818 |      0
 22 |   1328 |        1 |    311 |  46819 |      0
 23 |   1016 |        1 |    311 |  46820 |      0
 24 |    704 |        1 |    311 |  46821 |      0
 25 |    392 |        1 |    311 |  46822 |      0
(25 rows)

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 1));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46823 |      0
(1 row)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
            16384
(1 row)

testdb=# 

■「別ブロックへの更新処理」の何が問題か

• 空き領域のあるブロックを探す
• 空き領域のあるブロックが見つからない場合は、ファイルを拡張して新しいブロックを作る
• ブロックをディスクから共有バッファに読み込む
• 共有バッファへ読み込んだブロックに対して新しいレコードを書き込む

■強制的に空き領域を確保して更新（UPDATE）に備える

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46856 |      0
  2 |   7568 |        1 |    311 |  46857 |      0
  3 |   7256 |        1 |    311 |  46858 |      0
  4 |   6944 |        1 |    311 |  46859 |      0
  5 |   6632 |        1 |    311 |  46860 |      0
  6 |   6320 |        1 |    311 |  46861 |      0
  7 |   6008 |        1 |    311 |  46862 |      0
  8 |   5696 |        1 |    311 |  46863 |      0
  9 |   5384 |        1 |    311 |  46864 |      0
 10 |   5072 |        1 |    311 |  46865 |      0
 11 |   4760 |        1 |    311 |  46866 |      0
 12 |   4448 |        1 |    311 |  46867 |      0
 13 |   4136 |        1 |    311 |  46868 |      0
 14 |   3824 |        1 |    311 |  46869 |      0
 15 |   3512 |        1 |    311 |  46870 |      0
 16 |   3200 |        1 |    311 |  46871 |      0
 17 |   2888 |        1 |    311 |  46872 |      0
 18 |   2576 |        1 |    311 |  46873 |      0
 19 |   2264 |        1 |    311 |  46874 |      0
 20 |   1952 |        1 |    311 |  46875 |      0
 21 |   1640 |        1 |    311 |  46876 |      0
 22 |   1328 |        1 |    311 |  46877 |      0
 23 |   1016 |        1 |    311 |  46878 |      0
(23 rows)

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 1));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46879 |      0
  2 |   7568 |        1 |    311 |  46880 |      0
(2 rows)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
            16384
(1 row)

testdb=# 

testdb=# UPDATE t1 SET uname = 'update 101                                                                                                                                                                                                                                                                             ' WHERE uid=101;
UPDATE 1
testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 0));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46856 |  46881
  2 |   7568 |        1 |    311 |  46857 |      0
  3 |   7256 |        1 |    311 |  46858 |      0
  4 |   6944 |        1 |    311 |  46859 |      0
  5 |   6632 |        1 |    311 |  46860 |      0
  6 |   6320 |        1 |    311 |  46861 |      0
  7 |   6008 |        1 |    311 |  46862 |      0
  8 |   5696 |        1 |    311 |  46863 |      0
  9 |   5384 |        1 |    311 |  46864 |      0
 10 |   5072 |        1 |    311 |  46865 |      0
 11 |   4760 |        1 |    311 |  46866 |      0
 12 |   4448 |        1 |    311 |  46867 |      0
 13 |   4136 |        1 |    311 |  46868 |      0
 14 |   3824 |        1 |    311 |  46869 |      0
 15 |   3512 |        1 |    311 |  46870 |      0
 16 |   3200 |        1 |    311 |  46871 |      0
 17 |   2888 |        1 |    311 |  46872 |      0
 18 |   2576 |        1 |    311 |  46873 |      0
 19 |   2264 |        1 |    311 |  46874 |      0
 20 |   1952 |        1 |    311 |  46875 |      0
 21 |   1640 |        1 |    311 |  46876 |      0
 22 |   1328 |        1 |    311 |  46877 |      0
 23 |   1016 |        1 |    311 |  46878 |      0
 24 |    704 |        1 |    311 |  46881 |      0
(24 rows)

testdb=# SELECT lp,lp_off,lp_flags,lp_len,t_xmin,t_xmax FROM heap_page_items(get_raw_page('t1', 1));
 lp | lp_off | lp_flags | lp_len | t_xmin | t_xmax
----+--------+----------+--------+--------+--------
  1 |   7880 |        1 |    311 |  46879 |      0
  2 |   7568 |        1 |    311 |  46880 |      0
(2 rows)

testdb=# SELECT pg_relation_size('t1');
 pg_relation_size
------------------
            16384
(1 row)

testdb=# 

■まとめ

■「更新可能ビュー」とは

■更新可能ビューの制約・前提条件

• 単一のテーブル、または更新可能ビューのみをFROM句に持つこと。（結合をしていないこと）
• ビューの定義がWITH、DISTINCT、GROUP BY、HAVING、LIMIT、OFFSETを持たないこと。
• ビューの定義がUNION、INTERSECT、EXCEPTを持たないこと。
• ビューの定義のselectリストに出てくるカラムが、元テーブルのカラムをそのまま参照していること。expressionやリテラル、関数であってはならない。システムカラムであってもならない。
• 元テーブルのカラムが二回以上ビューの定義に出てこないこと。
• ビューがsecurity_barrier属性を持たないこと。

■更新可能ビューを使ってみる

[snaga@devsv02 updatableview]$ /tmp/pgsql/bin/psql testdb
psql (9.3devel)
Type "help" for help.

testdb=# CREATE TABLE t1 ( uid INTEGER PRIMARY KEY, uname TEXT NOT NULL );
CREATE TABLE
testdb=# CREATE VIEW v1 AS SELECT * FROM t1;
CREATE VIEW
testdb=# select * FROM t1;
 uid | uname
-----+-------
(0 rows)

testdb=# 

testdb=# insert into t1 values ( 101, 'Park Gyu-Ri' );
INSERT 0 1
testdb=# select * from t1;
 uid |    uname
-----+-------------
 101 | Park Gyu-Ri
(1 row)

testdb=# select * from v1;
 uid |    uname
-----+-------------
 101 | Park Gyu-Ri
(1 row)

testdb=# 

testdb=# insert into v1 values ( 102, 'Han Seung-Yeon' );
INSERT 0 1
testdb=# select * from t1;
 uid |     uname
-----+----------------
 101 | Park Gyu-Ri
 102 | Han Seung-Yeon
(2 rows)

testdb=# select * from v1;
 uid |     uname
-----+----------------
 101 | Park Gyu-Ri
 102 | Han Seung-Yeon
(2 rows)

testdb=# 

testdb=# explain insert into v1 values ( 103, 'Nicole' );
                   QUERY PLAN
------------------------------------------------
 Insert on t1  (cost=0.00..0.01 rows=1 width=0)
   ->  Result  (cost=0.00..0.01 rows=1 width=0)
(2 rows)

testdb=# 

testdb=# explain update v1 set uname = 'Nicole' where uid = 102;
                               QUERY PLAN
-------------------------------------------------------------------------
 Update on t1  (cost=0.00..8.27 rows=1 width=10)
   ->  Index Scan using t1_pkey on t1  (cost=0.00..8.27 rows=1 width=10)
         Index Cond: (uid = 102)
(3 rows)

testdb=#

■ビューのselectリストに含まれないカラム

testdb=# create view v2 as select uid from t1;
CREATE VIEW
testdb=# insert into v2 values ( 103 );
ERROR:  null value in column "uname" violates not-null constraint
DETAIL:  Failing row contains (103, null).
STATEMENT:  insert into v2 values ( 103 );
testdb=#

testdb=# ALTER TABLE t1 ALTER COLUMN uname SET DEFAULT '(noname)';
ALTER TABLE
testdb=# insert into v2 values ( 103 );
INSERT 0 1
testdb=# select * from t1;
 uid |     uname
-----+----------------
 101 | Park Gyu-Ri
 102 | Han Seung-Yeon
 103 | (noname)
(3 rows)

testdb=#

• 元テーブルのカラムにデフォルト値があればデフォルト値に設定。
• デフォルト値が無ければnullに設定。
• その上で、元テーブルの制約でチェックされる。

■まとめ

■単体テストツール「pgTAP」とは

■インストール方法

[snaga@devsv02 pgtap-0.90.0]$ ls
Changes  doc       META.json           pgtap-schema.control  src
compat   getos.sh  pgtap.control       README.md             test
contrib  Makefile  pgtap-core.control  sql                   tocgen
[snaga@devsv02 pgtap-0.90.0]$ env PATH=/usr/pgsql-9.1/bin:$PATH USE_PGXS=1 make
cp sql/pgtap.sql.in sql/pgtap.sql
sed -e 's,MODULE_PATHNAME,$libdir/pgtap,g' -e 's,__OS__,linux,g' -e 's,__VERSION__,0.90,g' sql/pgtap.sql > sql/pgtap.tmp
mv sql/pgtap.tmp sql/pgtap.sql
(...snip...)
cp sql/pgtap.sql sql/pgtap--0.90.0.sql
cp sql/pgtap-core.sql sql/pgtap-core--0.90.0.sql
cp sql/pgtap-schema.sql sql/pgtap-schema--0.90.0.sql
[snaga@devsv02 pgtap-0.90.0]$

[snaga@devsv02 pgtap-0.90.0]$ su
Password:
[root@devsv02 pgtap-0.90.0]# env PATH=/usr/pgsql-9.1/bin:$PATH USE_PGXS=1 make install
/bin/mkdir -p '/usr/pgsql-9.1/share/extension'
/bin/mkdir -p '/usr/share/doc/pgsql/extension'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 644 ./pgtap.control ./pgtap-core.control ./pgtap-schema.control '/usr/pgsql-9.1/share/extension/'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 644 ./sql/pgtap--0.90.0.sql ./sql/pgtap-core--0.90.0.sql ./sql/pgtap-schema--0.90.0.sql ./sql/pgtap--unpackaged--0.26.0.sql ./sql/pgtap--0.90.0.sql ./sql/pgtap-core--0.90.0.sql ./sql/pgtap-schema--0.90.0.sql  '/usr/pgsql-9.1/share/extension/'
/bin/sh /usr/pgsql-9.1/lib/pgxs/src/makefiles/../../config/install-sh -c -m 644 ./doc/pgtap.mmd '/usr/share/doc/pgsql/extension/'
[root@devsv02 pgtap-0.90.0]#

[snaga@devsv02 pgtap-0.90.0]$ psql -U postgres testdb
psql (9.1.2)
Type "help" for help.

testdb=# CREATE EXTENSION pgtap;
CREATE EXTENSION
testdb=# \q
[snaga@devsv02 pgtap-0.90.0]$

■単体テストの構造

BEGIN;
SELECT plan(13);

SELECT * FROM finish();
ROLLBACK;

BEGIN;
SELECT plan(1);

SELECT tables_are (
        'public',
        ARRAY[
        't1'
        ]
);

SELECT * FROM finish();
ROLLBACK;

[snaga@devsv02 t]$ psql -f t1.sql testdb
BEGIN
 plan
------
 1..1
(1 row)

                     tables_are
-----------------------------------------------------
 ok 1 - Schema public should have the correct tables
(1 row)

 finish
--------
(0 rows)

ROLLBACK
[snaga@devsv02 t]$

               finish
-------------------------------------
 # Looks like you failed 1 test of 1
(1 row)

■テーブルとカラムのテスト

SELECT tables_are (
        'pgperf',
        ARRAY[
        'snapshot',
        'snapshot_pg_stat_bgwriter'
        ]
);

                     tables_are
-----------------------------------------------------
 ok 1 - Schema pgperf should have the correct tables
(1 row)

                           tables_are
-----------------------------------------------------------------
 not ok 1 - Schema pgperf should have the correct tables        +
 # Failed test 1: "Schema pgperf should have the correct tables"+
 #     Missing tables:                                          +
 #         snapshot_pg_stat_user_tables                         +
 #         snapshot_pg_database_size
(1 row)

SELECT col_type_is('pgperf', 'snapshot_pg_stat_bgwriter', 'sid', 'integer', 'sid');
SELECT col_type_is('pgperf', 'snapshot_pg_stat_bgwriter', 'checkpoints_timed', 'bigint', 'checkpoints_timed');

 col_type_is
-------------
 ok 3 - sid
(1 row)

       col_type_is
-------------------------
 not ok 4 - sid         +
 # Failed test 4: "sid" +
 #         have: integer+
 #         want: bigint
(1 row)

                           col_type_is
------------------------------------------------------------------
 not ok 4 - sid2                                                 +
 # Failed test 4: "sid2"                                         +
 #    Column pgperf.snapshot_pg_stat_bgwriter.sid2 does not exist
(1 row)

■クエリ実行結果のテスト

testdb=> SELECT * FROM t1;
 uid |     uname
-----+----------------
 101 | Park Gyu-Ri
 102 | Han Seung-Yeon
 103 | Nicole
 104 | Koo Ha-Ra
 105 | Kang Ji-Young
(5 rows)

testdb=> SELECT uname FROM t1 WHERE uid=103;
 uname
--------
 Nicole
(1 row)

SELECT results_eq (
        'SELECT uname FROM t1 WHERE uid=103',
        ARRAY[ 'Nicole' ]
);

testdb=> SELECT uname FROM t1 WHERE uid=103 OR uid=104;
   uname
-----------
 Nicole
 Koo Ha-Ra
(2 rows)

SELECT results_eq (
        'SELECT uname FROM t1 WHERE uid=103 OR uid=104',
        ARRAY[ 'Nicole', 'Koo Ha-Ra' ]
);

testdb=> SELECT * FROM t1 WHERE uid=103;
 uid | uname
-----+--------
 103 | Nicole
(1 row)

SELECT results_eq (
        'SELECT * FROM t1 WHERE uid=103',
        $$ VALUES (103, 'Nicole') $$
);

testdb=> SELECT * FROM t1 WHERE uid=103 OR uid=104;
 uid |   uname
-----+-----------
 103 | Nicole
 104 | Koo Ha-Ra
(2 rows)

SELECT results_eq (
        'SELECT * FROM t1 WHERE uid=103 OR uid=104',
        $$ VALUES (103, 'Nicole'), (104, 'Koo Ha-Ra') $$
);

■プロシージャ、ユーザ定義関数のテスト

SELECT results_eq(
        'SELECT pgperf.create_snapshot_pg_stat_bgwriter(1)',
        ARRAY[ true ]
);

■pg_proveコマンドによる一括実行と集計

[snaga@devsv02 t]$ pg_prove -d testdb test_pg_stat_bgwriter.sql
t/test_pg_stat_bgwriter.sql .. ok
All tests successful.
Files=1, Tests=18,  0 wallclock secs ( 0.02 usr +  0.01 sys =  0.03 CPU)
Result: PASS
[snaga@devsv02 t]$ 

[snaga@devsv02 t]$ pg_prove -d testdb test_pg_stat_bgwriter.sql
t/test_pg_stat_bgwriter.sql .. 1/21
# Failed test 4: "sid"
#         have: integer
#         want: bigint
# Failed test 5: "sid2"
#    Column pgperf.snapshot_pg_stat_bgwriter.sid2 does not exist
# Looks like you failed 2 tests of 21
t/test_pg_stat_bgwriter.sql .. Failed 2/21 subtests

Test Summary Report
-------------------
t/test_pg_stat_bgwriter.sql (Wstat: 0 Tests: 21 Failed: 2)
  Failed tests:  4-5
Files=1, Tests=21,  0 wallclock secs ( 0.02 usr +  0.01 sys =  0.03 CPU)
Result: FAIL
[snaga@devsv02 t]$ 

■まとめ

■ブロックサイズの変更方法と確認方法

./configure --prefix=/usr/local/pgsql --with-blocksize=32

postgres=# show block_size;
 block_size
------------
 32768
(1 row)

postgres=#

■テスト環境とデータ

• NEC Express5800 GT110b
• Xeon Intel Xeon X3440 2.53GHz (1P4C)
• Unbeffered ECC 16GB
• Intel 520 Series 180GB
• Hitachi Deskstar 7K1000 HDS72101
• Red Hat Enterprise Linux 6.3 (x86_64)

/dev/mapper/vg_devsv03-lv_root on / type ext4 (rw)
/dev/mapper/vg_devsv03-lv_home on /home type ext4 (rw)
/dev/sda1 on /disk/disk1 type ext4 (rw,discard)

shared_buffers = 2048MB
wal_buffers = 32MB
checkpoint_segments = 128
checkpoint_timeout = 60min
autovacuum = off

■データのローディングとインデックスの作成

dbt3=# \d+
                          List of relations
 Schema |      Name       | Type  | Owner |    Size    | Description
--------+-----------------+-------+-------+------------+-------------
 public | customer        | table | snaga | 276 MB     |
 public | lineitem        | table | snaga | 8324 MB    |
 public | nation          | table | snaga | 8192 bytes |
 public | orders          | table | snaga | 1979 MB    |
 public | part            | table | snaga | 317 MB     |
 public | partsupp        | table | snaga | 1335 MB    |
 public | region          | table | snaga | 8192 bytes |
 public | supplier        | table | snaga | 17 MB      |
(8 rows)

dbt3=#

■テーブルフルスキャン

select count(*) from lineitem;

                                QUERY PLAN
---------------------------------------------------------------------------
 Aggregate  (cost=1279737.95..1279737.96 rows=1 width=0)
   ->  Seq Scan on lineitem  (cost=0.00..1129772.56 rows=59986156 width=0)
(2 rows)

dbt3=# \timing
Timing is on.
dbt3=# SELECT count(*) FROM lineitem;
  count
----------
 59986052
(1 row)

Time: 6457.410 ms
dbt3=# \i rpt_buf_cache.sql
   name   | num_buf | num_blks | pct
----------+---------+----------+------
 lineitem |       8 |  1057132 | 0.00
<others> |      46 |          |
(2 rows)

Time: 32.223 ms
dbt3=# SHOW block_size ;
 block_size
------------
 32768
(1 row)

Time: 0.223 ms
dbt3=#

■GROUP BYによる集約

select date_trunc('month', o_orderdate), sum(o_totalprice)
  from orders
 group by date_trunc('month', o_orderdate)
 order by 1;

                                    QUERY PLAN
----------------------------------------------------------------------------------
 Sort  (cost=363007.84..363013.85 rows=2406 width=8)
   Sort Key: (date_trunc('month'::text, (o_orderdate)::timestamp with time zone))
   ->  HashAggregate  (cost=362836.62..362872.71 rows=2406 width=8)
         ->  Seq Scan on orders  (cost=0.00..287837.71 rows=14999781 width=8)
(4 rows)

■結合と集約

select o.o_custkey,count(*)
  from orders o, lineitem l
 where l.l_commitdate < l.l_receiptdate
   and l.l_orderkey = o.o_orderkey
 group by o.o_custkey
 order by 2 desc limit 20;

                                               QUERY PLAN
--------------------------------------------------------------------------------------------------------
 Limit  (cost=7004441.17..7004441.22 rows=20 width=4)
   ->  Sort  (cost=7004441.17..7006622.04 rows=872351 width=4)
         Sort Key: (count(*))
         ->  GroupAggregate  (cost=6822539.33..6981228.22 rows=872351 width=4)
               ->  Sort  (cost=6822539.33..6872527.79 rows=19995384 width=4)
                     Sort Key: o.o_custkey
                     ->  Hash Join  (cost=649382.08..3304284.73 rows=19995384 width=4)
                           Hash Cond: (l.l_orderkey = o.o_orderkey)
                           ->  Seq Scan on lineitem l  (cost=0.00..1815236.90 rows=19995384 width=4)
                                 Filter: (l_commitdate < l_receiptdate)
                           ->  Hash  (cost=403281.59..403281.59 rows=15000359 width=8)
                                 ->  Seq Scan on orders o  (cost=0.00..403281.59 rows=15000359 width=8)
(12 rows)

                                                        QUERY PLAN
--------------------------------------------------------------------------------------------------------------------------
 Limit  (cost=4970682.69..4970682.74 rows=20 width=4)
   ->  Sort  (cost=4970682.69..4972685.24 rows=801019 width=4)
         Sort Key: (count(*))
         ->  GroupAggregate  (cost=4791392.62..4949367.86 rows=801019 width=4)
               ->  Sort  (cost=4791392.62..4841380.97 rows=19995341 width=4)
                     Sort Key: o.o_custkey
                     ->  Merge Join  (cost=44.76..2093263.54 rows=19995341 width=4)
                           Merge Cond: (o.o_orderkey = l.l_orderkey)
                           ->  Index Scan using pk_orders on orders o  (cost=0.00..328663.08 rows=15000450 width=8)
                           ->  Index Scan using i_l_orderkey on lineitem l  (cost=0.00..1477217.99 rows=19995341 width=4)
                                 Filter: (l_commitdate < l_receiptdate)
(11 rows)

■まとめ

■「テーブルパーティショニング」

■PostgreSQLのテーブルパーティショニング（Constraint Exclusion）

■パーティションの作成

CREATE TABLE orders (
        o_orderkey INTEGER PRIMARY KEY,
        o_custkey INTEGER,
        o_orderstatus CHAR(1),
        o_totalprice REAL,
        o_orderdate DATE,
        o_orderpriority CHAR(15),
        o_clerk CHAR(15),
        o_shippriority INTEGER,
        o_comment VARCHAR(79)
);

testdb=# CREATE TABLE orders (
testdb(#         o_orderkey INTEGER PRIMARY KEY,
testdb(#         o_custkey INTEGER,
testdb(#         o_orderstatus CHAR(1),
testdb(#         o_totalprice REAL,
testdb(#         o_orderDATE DATE,
testdb(#         o_orderpriority CHAR(15),
testdb(#         o_clerk CHAR(15),
testdb(#         o_shippriority INTEGER,
testdb(#         o_comment VARCHAR(79)
testdb(# );
NOTICE:  CREATE TABLE / PRIMARY KEY will create implicit index "orders_pkey" for table "orders"
CREATE TABLE
testdb=# CREATE INDEX orders_o_orderdate_idx ON orders(o_orderdate);
CREATE INDEX
testdb=# \d orders
                Table "public.orders"
     Column      |         Type          | Modifiers
-----------------+-----------------------+-----------
 o_orderkey      | integer               | not null
 o_custkey       | integer               |
 o_orderstatus   | character(1)          |
 o_totalprice    | real                  |
 o_orderdate     | date                  |
 o_orderpriority | character(15)         |
 o_clerk         | character(15)         |
 o_shippriority  | integer               |
 o_comment       | character varying(79) |
Indexes:
    "orders_pkey" PRIMARY KEY, btree (o_orderkey)
    "orders_o_orderdate_idx" btree (o_orderdate)

testdb=#

testdb=# CREATE TABLE orders_1992 () INHERITS (orders);
CREATE TABLE
testdb=# \d
            List of relations
 Schema |    Name     | Type  |  Owner
--------+-------------+-------+----------
 public | orders      | table | postgres
 public | orders_1992 | table | postgres
(2 rows)

testdb=# \d orders_1992
             Table "public.orders_1992"
     Column      |         Type          | Modifiers
-----------------+-----------------------+-----------
 o_orderkey      | integer               | not null
 o_custkey       | integer               |
 o_orderstatus   | character(1)          |
 o_totalprice    | real                  |
 o_orderdate     | date                  |
 o_orderpriority | character(15)         |
 o_clerk         | character(15)         |
 o_shippriority  | integer               |
 o_comment       | character varying(79) |
Inherits: orders

testdb=#

testdb=# ALTER TABLE orders_1992 ADD PRIMARY KEY (o_orderkey);
NOTICE:  ALTER TABLE / ADD PRIMARY KEY will create implicit index "orders_1992_pkey" for table "orders_1992"
ALTER TABLE
testdb=# 

testdb=# ALTER TABLE orders_1992 ADD CHECK(o_orderdate >= '1992-01-01' AND o_orderdate < '1993-01-01');
ALTER TABLE
testdb=# \d orders_1992
             Table "public.orders_1992"
     Column      |         Type          | Modifiers
-----------------+-----------------------+-----------
 o_orderkey      | integer               | not null
 o_custkey       | integer               |
 o_orderstatus   | character(1)          |
 o_totalprice    | real                  |
 o_orderdate     | date                  |
 o_orderpriority | character(15)         |
 o_clerk         | character(15)         |
 o_shippriority  | integer               |
 o_comment       | character varying(79) |
Indexes:
    "orders_1992_pkey" PRIMARY KEY, btree (o_orderkey)
Check constraints:
    "orders_1992_o_orderdate_check" CHECK (o_orderdate >= '1992-01-01'::date AND o_orderdate < '1993-01-01'::date)
Inherits: orders2

testdb=#

testdb=# \d
            List of relations
 Schema |    Name     | Type  |  Owner
--------+-------------+-------+----------
 public | orders      | table | postgres
 public | orders_1992 | table | postgres
 public | orders_1993 | table | postgres
 public | orders_1994 | table | postgres
 public | orders_1995 | table | postgres
 public | orders_1996 | table | postgres
 public | orders_1997 | table | postgres
 public | orders_1998 | table | postgres
(8 rows)

testdb=#

■パーティションへのデータのローディング

• 親テーブルにINSERTトリガを設定し、親テーブルへのINSERTを適切なパーティション（子テーブル）に自動的に振り分ける方法
• パーティションを直接指定してデータをローディングする方法

■パーティションテーブルに対する実行プラン

SELECT sum(o_totalprice) FROM orders;

testdb=# EXPLAIN
SELECT sum(o_totalprice) FROM orders;
                              QUERY PLAN
----------------------------------------------------------------------
 Aggregate  (cost=44076.00..44076.01 rows=1 width=4)
   ->  Seq Scan on orders  (cost=0.00..40326.00 rows=1500000 width=4)
(2 rows)

testdb=#

testdb=# EXPLAIN SELECT sum(o_totalprice) FROM orders2;
                                      QUERY PLAN

--------------------------------------------------------------------------------
-------
 Aggregate  (cost=44079.01..44079.02 rows=1 width=4)
   ->  Append  (cost=0.00..40329.00 rows=1500001 width=4)
         ->  Seq Scan on orders2  (cost=0.00..0.00 rows=1 width=4)
         ->  Seq Scan on orders_1992 orders2  (cost=0.00..6104.89 rows=227089 width=4)
         ->  Seq Scan on orders_1993 orders2  (cost=0.00..6093.45 rows=226645 width=4)
         ->  Seq Scan on orders_1994 orders2  (cost=0.00..6117.97 rows=227597 width=4)
         ->  Seq Scan on orders_1995 orders2  (cost=0.00..6146.37 rows=228637 width=4)
         ->  Seq Scan on orders_1996 orders2  (cost=0.00..6148.26 rows=228626 width=4)
         ->  Seq Scan on orders_1997 orders2  (cost=0.00..6124.83 rows=227783 width=4)
         ->  Seq Scan on orders_1998 orders2  (cost=0.00..3593.23 rows=133623 width=4)
(10 rows)

testdb=# 

■テーブルパーティショニングによるパフォーマンス向上

SELECT sum(o_totalprice) FROM orders
 WHERE o_orderdate >= '1993-09-01'
   AND o_orderdate <= '1993-09-30';

testdb=# EXPLAIN SELECT sum(o_totalprice) FROM orders
 WHERE o_orderdate >= '1993-09-01'
   AND o_orderdate <= '1993-09-30';
                                          QUERY PLAN
-----------------------------------------------------------------------------------------------
 Aggregate  (cost=33690.55..33690.56 rows=1 width=4)
   ->  Seq Scan on orders  (cost=0.00..33683.58 rows=2786 width=4)
         Filter: ((o_orderdate >= '1993-09-01'::date) AND (o_orderdate <= '1993-09-30'::date))
(3 rows)

testdb=# 

testdb=# EXPLAIN SELECT sum(o_totalprice) FROM orders
 WHERE o_orderdate >= '1993-09-01'
   AND o_orderdate <= '1993-09-30';
                                             QUERY PLAN
-----------------------------------------------------------------------------------------------------
 Aggregate  (cost=7270.49..7270.50 rows=1 width=4)
   ->  Append  (cost=0.00..7226.67 rows=17525 width=4)
         ->  Seq Scan on orders  (cost=0.00..0.00 rows=1 width=4)
               Filter: ((o_orderdate >= '1993-09-01'::date) AND (o_orderdate <= '1993-09-30'::date))
         ->  Seq Scan on orders_1993 orders  (cost=0.00..7226.67 rows=17524 width=4)
               Filter: ((o_orderdate >= '1993-09-01'::date) AND (o_orderdate <= '1993-09-30'::date))
(6 rows)

testdb=#

testdb=# SELECT sum(o_totalprice) FROM orders
 WHERE o_orderdate >= '1993-09-01'
   AND o_orderdate <= '1993-09-30';
    sum
------------
 2.8482e+09
(1 row)

Time: 405.432 ms
testdb=# 

testdb=# SELECT sum(o_totalprice) FROM orders
 WHERE o_orderdate >= '1993-09-01'
   AND o_orderdate <= '1993-09-30';
    sum
------------
 2.8482e+09
(1 row)

Time: 72.615 ms
testdb=#

■まとめ

■テーブルパーティションを操作するpg_partパッケージ

• パーティションの作成（＋データの移行）
• パーティションの解除（＋データの移行）
• パーティションの追加（アタッチ）
• パーティションの切り離し（デタッチ）

[snaga@devsv03 ~]$ psql -f pg_part.sql dbt3
BEGIN
CREATE FUNCTION
CREATE FUNCTION
(...snip...)
CREATE FUNCTION
CREATE FUNCTION
COMMIT
[snaga@devsv03 ~]$

■パーティションの作成

SELECT pgpart.add_partition(
  schema_name,
  table_name,
  partition_name,
  check_condition,
  temp_file);

dbt3=# SELECT pgpart.add_partition(
dbt3(#   'public',
dbt3(#   'orders',
dbt3(#   'orders_1992',
dbt3(#   ' ''1992-01-01'' <= o_orderdate AND o_orderdate < ''1993-01-01'' ',
dbt3(#   '/tmp/orders.tmp');
 add_partition
---------------
 t
(1 row)

dbt3=#

• 親テーブルを継承して子テーブル（パーティション）を作成
• 親テーブルからパーティションに移動させるレコードをエクスポート
• 親テーブルからパーティションに移動させるレコードを削除
• エクスポートしておいたレコードをパーティションにインポート
• パーティションに親テーブルと同等の主キー制約を追加
• パーティションに親テーブルと同等のインデックスを追加

psql:add_part.sql:8: NOTICE:  add_partition: CREATE TABLE public.orders_1992( CONSTRAINT __orders_1992_check CHECK( '1992-01-01' <= o_orderdate AND o_orderdate < '1993-01-01' )) INHERITS (public.orders);
psql:add_part.sql:8: NOTICE:  add_partition: COPY ( SELECT * FROM public.orders WHERE  '1992-01-01' <= o_orderdate AND o_orderdate < '1993-01-01'  ) to '/tmp/orders.tmp';
psql:add_part.sql:8: NOTICE:  add_partition: DELETE FROM public.orders WHERE  '1992-01-01' <= o_orderdate AND o_orderdate < '1993-01-01' ;
psql:add_part.sql:8: NOTICE:  add_partition: COPY public.orders_1992 FROM '/tmp/orders.tmp';
psql:add_part.sql:8: NOTICE:  add_partition: ALTER TABLE public.orders_1992 ADD PRIMARY KEY (o_orderkey);
psql:add_part.sql:8: NOTICE:  add_partition: CREATE INDEX orders_1992_o_orderdate_idx ON public.orders_1992 USING btree (o_orderdate);
psql:add_part.sql:8: NOTICE:  add_partition: CREATE INDEX orders_1992_o_custkey_idx ON public.orders_1992 USING btree (o_custkey);

■パーティション一覧の取得

SELECT pgpart.show_partition(schema_name, table_name);

dbt3=# SELECT pgpart.show_partition('public', 'orders');
 show_partition
----------------
 orders_1992
 orders_1993
 orders_1994
 orders_1995
 orders_1996
 orders_1997
(6 rows)

dbt3=#

■パーティションのマージ（解除）

SELECT pgpart.merge_partition(
  schema_name,
  table_name,
  partition_name,
  check_constraint,
  temp_file);

dbt3=# SELECT pgpart.merge_partition('public', 'orders', 'orders_1992', null, '/tmp/orders.tmp');
 merge_partition
-----------------
 t
(1 row)

dbt3=# 

■パーティションのアタッチ（追加）

SELECT pgpart.attach_partition (
  schema_name,
  table_name,
  partition_name,
  check_condition);

dbt3=# SELECT pgpart.show_partition('public', 'orders');
 show_partition
----------------
 orders_1992
 orders_1993
 orders_1994
 orders_1995
 orders_1996
 orders_1997
(6 rows)

dbt3=# SELECT pgpart.attach_partition('public', 'orders', 'orders_1998', ' ''1998-01-01'' <= o_orderdate AND o_orderdate < ''1999-01-01'' ');
 attach_partition
------------------
 t
(1 row)

dbt3=# SELECT pgpart.show_partition('public', 'orders');
 show_partition
----------------
 orders_1992
 orders_1993
 orders_1994
 orders_1995
 orders_1996
 orders_1997
 orders_1998
(7 rows)

dbt3=# 

dbt3=# EXPLAIN SELECT count(*) FROM orders WHERE o_orderdate = '1998-01-01';
                                                QUERY PLAN
----------------------------------------------------------------------------------------------------------
 Aggregate  (cost=6044.09..6044.10 rows=1 width=0)
   ->  Append  (cost=0.00..6027.42 rows=6667 width=0)
         ->  Seq Scan on orders  (cost=0.00..0.00 rows=1 width=0)
               Filter: (o_orderdate = '1998-01-01'::date)
         ->  Bitmap Heap Scan on orders_1998 orders  (cost=72.27..6027.42 rows=6666 width=0)
               Recheck Cond: (o_orderdate = '1998-01-01'::date)
               ->  Bitmap Index Scan on orders_1998_o_orderdate_idx  (cost=0.00..70.61 rows=6666 width=0)
                     Index Cond: (o_orderdate = '1998-01-01'::date)
(8 rows)

dbt3=#

■パーティションのデタッチ（削除）

SELECT pgpart.detach_partition (
  schema_name,
  table_name,
  partition_name);

dbt3=# SELECT pgpart.show_partition('public', 'orders');
 show_partition
----------------
 orders_1992
 orders_1993
 orders_1994
 orders_1995
 orders_1996
 orders_1997
 orders_1998
(7 rows)

dbt3=# SELECT pgpart.detach_partition('public', 'orders', 'orders_1992');
 detach_partition
------------------
 t
(1 row)

dbt3=# SELECT pgpart.show_partition('public', 'orders');
 show_partition
----------------
 orders_1993
 orders_1994
 orders_1995
 orders_1996
 orders_1997
 orders_1998
(6 rows)

dbt3=#

■パーティション作成時の手順

dbt3=# \d+
                        List of relations
 Schema |      Name       | Type  | Owner |  Size   | Description
--------+-----------------+-------+-------+---------+-------------
 public | orders          | table | snaga | 1964 MB |
 public | orders_1992     | table | snaga | 299 MB  |
 public | orders_1993     | table | snaga | 298 MB  |
(...snip...)

dbt3=# VACUUM orders;
VACUUM
dbt3=# \d+
                        List of relations
 Schema |      Name       | Type  | Owner |  Size   | Description
--------+-----------------+-------+-------+---------+-------------
 public | orders          | table | snaga | 0 bytes |
 public | orders_1992     | table | snaga | 299 MB  |
 public | orders_1993     | table | snaga | 298 MB  |
(...snip...)

dbt3=#

■まとめ

■「MPP」とは何か

■PostgreSQL用MPPミドルウェア「Stado」

■導入手順

• Stadoのインストール
• PostgreSQLのセットアップ
• Stadoのセットアップ
• ユーザデータベースの作成
• テーブルスペースへのパーティションの移動
• テーブルの作成とデータのロード

■導入環境

• NEC Express5800 GT110b
• Xeon Intel Xeon X3440 2.53GHz (1P4C)
• Unbeffered ECC 16GB
• Hitachi Deskstar 7K1000 HDS72101
• Red Hat Enterprise Linux 6.3 (x86_64)

[snaga@devsv03 ~]$ rpm -qa | grep postgresql
postgresql92-9.2.2-1PGDG.rhel6.x86_64
postgresql92-devel-9.2.2-1PGDG.rhel6.x86_64
postgresql92-server-9.2.2-1PGDG.rhel6.x86_64
postgresql92-contrib-9.2.2-1PGDG.rhel6.x86_64
postgresql92-libs-9.2.2-1PGDG.rhel6.x86_64
[snaga@devsv03 ~]$ rpm -qa | grep openjdk
java-1.6.0-openjdk-devel-1.6.0.0-1.45.1.11.1.el6.x86_64
java-1.6.0-openjdk-1.6.0.0-1.45.1.11.1.el6.x86_64
[snaga@devsv03 ~]$

■Stadoのインストール

[snaga@devvm03 tmp]$ bzr branch lp:~sgdg/stado/stado
You have not informed bzr of your Launchpad ID, and you must do this to
write to Launchpad or access private data.  See "bzr help launchpad-login".
Branched 55 revision(s).
[snaga@devvm03 tmp]$ 

[snaga@devsv03 stado]$ wget http://www.uptime.jp/downloads/stado/stado-20121223.tar.gz
(...snip...)

[snaga@devsv03 stado]$ wget http://www.uptime.jp/downloads/stado/install_stado.sh
(...snip...)

[snaga@devsv03 stado]$ ls
install_stado.sh  stado-20121223.tar.gz
[snaga@devsv03 stado]$ tar zxf stado-20121223.tar.gz
[snaga@devsv03 stado]$ ls
install_stado.sh  stado-20121223/  stado-20121223.tar.gz
[snaga@devsv03 stado]$ cd stado-20121223
[snaga@devsv03 stado-20121223]$ ls -F
bin/    build.xml  docs/  lib/   README.TXT  stado.config
build/  dist/      jars/  misc/  src/
[snaga@devsv03 stado-20121223]$ su
Password:
[root@devsv03 stado-20121223]# sh ../install_stado.sh
[root@devsv03 stado-20121223]# ls -l /usr/local/stado/
total 16
drwxr-xr-x. 2 stado stado 4096 Dec 23 18:06 bin/
drwxr-xr-x. 2 stado stado 4096 Dec 23 18:06 config/
drwxr-xr-x. 2 stado stado 4096 Dec 23 18:06 lib/
drwxrwxr-x. 2 stado stado 4096 Dec 23 18:06 log/
[root@devsv03 stado-20121223]#

■PostgreSQLのセットアップ

listen_addresses = '*'
max_connections = 100
shared_buffers = 2GB
work_mem = 2GB
maintenance_work_mem = 2GB
wal_buffers = 32MB
checkpoint_segments = 128
checkpoint_timeout = 60min
log_filename = 'postgresql-%Y%m%d.log'
log_min_duration_statement = 0
log_checkpoints = on
log_connections = on
log_disconnections = on
log_line_prefix = '[%t] %p: '
log_temp_files = 0
autovacuum = off

[root@devsv03 pgsql]# df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/sda3            936146792 739057756 148768328  84% /
/dev/sda1               101086     12010     83857  13% /boot
tmpfs                  8213720         0   8213720   0% /dev/shm
/dev/sdb1            961432072 135545524 777048548  15% /disk/disk2
/dev/sdc1            961432072  53769008 858825064   6% /disk/disk3
/dev/sdd1            961432072 134776980 777817092  15% /disk/disk4
[root@devsv03 pgsql]# 
[root@devsv03 pgsql]# mkdir /disk/disk2/pgsql
[root@devsv03 pgsql]# mkdir /disk/disk3/pgsql
[root@devsv03 pgsql]# mkdir /disk/disk4/pgsql
[root@devsv03 pgsql]# chown postgres:postgres /disk/disk4/pgsql /disk/disk3/pgsql /disk/disk2/pgsql

postgres=# select * from pg_tablespace;
  spcname   | spcowner | spcacl | spcoptions
------------+----------+--------+------------
 pg_default |       10 |        |
 pg_global  |       10 |        |
(2 rows)

postgres=# create tablespace tblspc2 location '/disk/disk2/pgsql';
CREATE TABLESPACE
postgres=# create tablespace tblspc3 location '/disk/disk3/pgsql';
CREATE TABLESPACE
postgres=# create tablespace tblspc4 location '/disk/disk4/pgsql';
CREATE TABLESPACE
postgres=# select * from pg_tablespace;
  spcname   | spcowner | spcacl | spcoptions
------------+----------+--------+------------
 pg_default |       10 |        |
 pg_global  |       10 |        |
 tblspc2    |       10 |        |
 tblspc3    |       10 |        |
 tblspc4    |       10 |        |
(5 rows)

postgres=#

[root@devsv03 pgsql]# ls -l /var/lib/pgsql/9.2/data/pg_tblspc
total 0
lrwxrwxrwx. 1 postgres postgres 17 Dec 23 18:38 16384 -> /disk/disk2/pgsql
lrwxrwxrwx. 1 postgres postgres 17 Dec 23 18:38 16385 -> /disk/disk3/pgsql
lrwxrwxrwx. 1 postgres postgres 17 Dec 23 18:38 16386 -> /disk/disk4/pgsql
[root@devsv03 pgsql]#

[stado@devsv03 ~]$ createuser -d -E -U postgres -P stado
Enter password for new role:
Enter it again:
[stado@devsv03 ~]$ psql -h localhost -U stado postgres
psql (9.2.2)
Type "help" for help.

postgres=> \q
[stado@devsv03 ~]$

■Stadoのセットアップ

xdb.coordinator.host=127.0.0.1

xdb.port=6453
xdb.maxconnections=10

xdb.default.dbusername=stado
xdb.default.dbpassword=stado

xdb.default.dbport=5432

xdb.default.threads.pool.initsize=2
xdb.default.threads.pool.maxsize=10

xdb.metadata.database=XDBSYS
xdb.metadata.dbhost=127.0.0.1

xdb.nodecount=4

xdb.node.1.dbhost=127.0.0.1
xdb.node.2.dbhost=127.0.0.1
xdb.node.3.dbhost=127.0.0.1
xdb.node.4.dbhost=127.0.0.1

xdb.coordinator.node=1

[stado@devsv03 ~]$ cd /usr/local/stado/bin
[stado@devsv03 bin]$ ./gs-createmddb.sh -u stadoadm -p password
Executed Statement: create table xsystablespaces ( tablespaceid serial, tablespacename varchar(255) not null, ownerid int not null, primary key(tablespaceid))
Executed Statement: create unique index idx_xsystablespaces_1 on xsystablespaces (tablespacename)
(...snip...)
Executed Statement: create unique index idx_xsyschecks_1 on xsyschecks (constid, seqno)
Executed Statement: alter table xsyschecks add foreign key (constid) references xsysconstraints (constid)
User stadoadm is created
[stado@devsv03 bin]$

■コーディネータプロセスの起動

[stado@devsv03 bin]$ ./gs-server.sh
Starting....
[stado@devsv03 bin]$ ps -aef | grep stado
root     29892 29721  0 18:59 pts/3    00:00:00 su - stado
stado    29893 29892  0 18:59 pts/3    00:00:00 -bash
stado    30045     1  2 19:06 pts/3    00:00:00 java -classpath ../lib/stado.jar:../lib/log4j.jar:../lib/postgresql.jar: -Xms512M -Xmx512M -Xss256K -Dconfig.file.path=../config/stado.config org.postgresql.stado.util.XdbServer
postgres 30064 29604  0 19:06 ?        00:00:00 postgres: stado XDBSYS 127.0.0.1(48401) idle
stado    30099 29893  1 19:06 pts/3    00:00:00 ps -aef
stado    30100 29893  0 19:06 pts/3    00:00:00 grep stado
[stado@devsv03 bin]$

■エージェントプロセスの起動

[stado@devsv03 bin]$ ./gs-agent.sh -n 4
Starting....
[stado@devsv03 bin]$ ps -aef | grep stado
root     29892 29721  0 18:59 pts/3    00:00:00 su - stado
stado    29893 29892  0 18:59 pts/3    00:00:00 -bash
stado    30045     1  0 19:06 pts/3    00:00:00 java -classpath ../lib/stado.jar:../lib/log4j.jar:../lib/postgresql.jar: -Xms512M -Xmx512M -Xss256K -Dconfig.file.path=../config/stado.config org.postgresql.stado.util.XdbServer
postgres 30064 29604  0 19:06 ?        00:00:00 postgres: stado XDBSYS 127.0.0.1(48401) idle
stado    30112     1  4 19:07 pts/3    00:00:00 java -classpath ../lib/stado.jar:../lib/log4j.jar:../lib/postgresql.jar: -Xms256M -Xmx256M -Dconfig.file.path=../config/stado_agent.config org.postgresql.stado.util.XdbAgent -n 4
stado    30135 29893  1 19:07 pts/3    00:00:00 ps -aef
stado    30136 29893  0 19:07 pts/3    00:00:00 grep stado
[stado@devsv03 bin]$

■ユーザーデータベースの作成

[stado@devsv03 bin]$ ./gs-createdb.sh -d testdb -u stadoadm -p password -n 1,2,3,4
OK
[stado@devsv03 bin]$ 

[stado@devsv03 bin]$ ./gs-cmdline.sh -d testdb -u stadoadm -p password

Stado -> show databases;
+------------------------------+
| DATABASE |  STATUS |  NODES  |
+------------------------------+
| testdb   | Started | 1,2,3,4 |
+------------------------------+
1 row(s).

Stado ->
[stado@devsv03 bin]$ 

■パーティションのテーブルスペースへの移動

postgres=> SELECT datname,dattablespace,spcname
postgres->   FROM pg_database d LEFT OUTER JOIN pg_tablespace s
postgres->        ON d.dattablespace=s.oid;
   datname    | dattablespace |  spcname
--------------+---------------+------------
 template1    |          1663 | pg_default
 template0    |          1663 | pg_default
 postgres     |          1663 | pg_default
 XDBSYS       |          1663 | pg_default
 __testdb__N1 |          1663 | pg_default
 __testdb__N2 |          1663 | pg_default
 __testdb__N3 |          1663 | pg_default
 __testdb__N4 |          1663 | pg_default
(8 rows)

postgres=>

[stado@devsv03 bin]$ ./gs-dbstop.sh -u stadoadm -p password -d testdb
Database(s) testdb stopped.
[stado@devsv03 bin]$ 

[stado@devsv03 bin]$ psql -U postgres postgres
psql (9.2.2)
Type "help" for help.

postgres=# ALTER DATABASE "__testdb__N2" SET TABLESPACE tblspc2;
ALTER DATABASE
postgres=# ALTER DATABASE "__testdb__N3" SET TABLESPACE tblspc3;
ALTER DATABASE
postgres=# ALTER DATABASE "__testdb__N4" SET TABLESPACE tblspc4;
ALTER DATABASE
postgres=# SELECT datname,dattablespace,spcname
postgres-#   FROM pg_database d LEFT OUTER JOIN pg_tablespace s
postgres-#        ON d.dattablespace=s.oid;
   datname    | dattablespace |  spcname
--------------+---------------+------------
 template1    |          1663 | pg_default
 template0    |          1663 | pg_default
 postgres     |          1663 | pg_default
 XDBSYS       |          1663 | pg_default
 __testdb__N1 |          1663 | pg_default
 __testdb__N2 |         16384 | tblspc2
 __testdb__N3 |         16385 | tblspc3
 __testdb__N4 |         16386 | tblspc4
(8 rows)

postgres=# \q
[stado@devsv03 bin]$ ./gs-dbstart.sh -u stadoadm -p password -d testdb
Database(s) testdb started.
[stado@devsv03 bin]$

■ユーザデータベースへの接続

[stado@devsv03 bin]$ psql -p 6453 -h localhost -U stadoadm testdb
Password for user stadoadm:
psql (9.2.2, server 9.0.1)
WARNING: psql version 9.2, server version 9.0.
         Some psql features might not work.
Type "help" for help.

testdb=>

■テーブルの作成とデータのロード

CREATE TABLE orders (
        o_orderkey INTEGER,
        o_custkey INTEGER,
        o_orderstatus CHAR(1),
        o_totalprice REAL,
        o_orderDATE DATE,
        o_orderpriority CHAR(15),
        o_clerk CHAR(15),
        o_shippriority INTEGER,
        o_comment VARCHAR(79)
)
REPLICATED;

CREATE TABLE orders2 (
        o_orderkey INTEGER,
        o_custkey INTEGER,
        o_orderstatus CHAR(1),
        o_totalprice REAL,
        o_orderDATE DATE,
        o_orderpriority CHAR(15),
        o_clerk CHAR(15),
        o_shippriority INTEGER,
        o_comment VARCHAR(79)
)
PARTITIONING KEY o_orderkey ON ALL;

testdb=> \i create_tables.sql
CREATE TABLE
CREATE TABLE
CREATE TABLE
CREATE TABLE
testdb=> \d+
                     List of relations
 Schema |   Name    | Type  | Owner |  Size   | Description
--------+-----------+-------+-------+---------+-------------
 public | lineitem  | table | stado | 0 bytes |
 public | lineitem2 | table | stado | 0 bytes |
 public | orders    | table | stado | 0 bytes |
 public | orders2   | table | stado | 0 bytes |
(4 rows)

testdb=>

[stado@devsv03 ~]$ cd /usr/local/stado/bin/
[stado@devsv03 bin]$ ./gs-cmdline.sh -u stadoadm -p password -d testdb

Stado -> show tables;
+-----------------------------------------------------+
|   TABLE   | TABLE_PARTITIONING_COLUMN | TABLE_NODES |
+-----------------------------------------------------+
| lineitem  |                           | 1,2,3,4     |
| lineitem2 | l_orderkey                | 1,2,3,4     |
| orders    |                           | 1,2,3,4     |
| orders2   | o_orderkey                | 1,2,3,4     |
+-----------------------------------------------------+
4 row(s).

Stado ->

testdb=> \i load_tables.sql
Timing is on.
Time: 162623.103 ms
Time: 64810.198 ms
Time: 763729.597 ms
Time: 320365.227 ms
testdb=> \d+
                     List of relations
 Schema |   Name    | Type  | Owner |  Size   | Description
--------+-----------+-------+-------+---------+-------------
 public | lineitem  | table | stado | 8326 MB |
 public | lineitem2 | table | stado | 2081 MB |
 public | orders    | table | stado | 1979 MB |
 public | orders2   | table | stado | 495 MB  |
(4 rows)

testdb=>

■クエリの実行

testdb=> select count(*) from orders;
 count(*)
----------
 15000000
(1 row)

Time: 1870.665 ms
testdb=> select count(*) from orders2;
 count(*)
----------
 15000000
(1 row)

Time: 650.026 ms
testdb=>

■まとめ

■参考資料

■マテリアライズドビューとは

■マテリアライズドビューを作成する

select c_name,
           date_trunc('month', o_orderdate),
           sum(o_totalprice)::numeric
      from orders o, customer c
     where o.o_custkey = c.c_custkey
     group by 1, 2;

dbt3=# explain select c_name,
           date_trunc('month', o_orderdate),
           sum(o_totalprice)::numeric
      from orders o, customer c
     where o.o_custkey = c.c_custkey
     group by 1, 2;
                                             QUERY PLAN
----------------------------------------------------------------------------------------------------
 GroupAggregate  (cost=360805.98..402055.98 rows=1500000 width=27)
   ->  Sort  (cost=360805.98..364555.98 rows=1500000 width=27)
         Sort Key: c.c_name, (date_trunc('month'::text, (o.o_orderdate)::timestamp with time zone))
         ->  Hash Join  (cost=7785.00..99265.00 rows=1500000 width=27)
               Hash Cond: (o.o_custkey = c.c_custkey)
               ->  Seq Scan on orders o  (cost=0.00..40326.00 rows=1500000 width=12)
               ->  Hash  (cost=5031.00..5031.00 rows=150000 width=23)
                     ->  Seq Scan on customer c  (cost=0.00..5031.00 rows=150000 width=23)
(8 rows)

dbt3=#

dbt3=# create materialized view customer_monthly_totalprice
dbt3-#  as select c_name,
dbt3-#            date_trunc('month', o_orderdate),
dbt3-#            sum(o_totalprice)::numeric
dbt3-#       from orders o, customer c
dbt3-#      where o.o_custkey = c.c_custkey
dbt3-#      group by 1, 2;
SELECT 1353175
dbt3=# 

dbt3=# select relname,relkind from pg_class where relname like 'cust%';
           relname           | relkind
-----------------------------+---------
 customer                    | r
 customer_monthly_totalprice | m
(2 rows)

dbt3=#

■マテリアライズドビューを使って検索する

dbt3=# explain select * from customer_monthly_totalprice;
                                      QUERY PLAN
--------------------------------------------------------------------------------------
 Seq Scan on customer_monthly_totalprice  (cost=0.00..25188.75 rows=1353175 width=34)
(1 row)

dbt3=#

■マテリアライズドビューを更新する

dbt3=# select * from customer_monthly_totalprice where c_name='Customer#000000007';
       c_name       |       date_trunc       |   sum
--------------------+------------------------+---------
 Customer#000000007 | 1992-03-01 00:00:00+00 |  322432
 Customer#000000007 | 1992-04-01 00:00:00+00 |  434825
 Customer#000000007 | 1993-06-01 00:00:00+00 |  501704
 Customer#000000007 | 1993-11-01 00:00:00+00 | 80777.5
 Customer#000000007 | 1994-02-01 00:00:00+00 |  176604
 Customer#000000007 | 1994-12-01 00:00:00+00 |  192318
 Customer#000000007 | 1995-09-01 00:00:00+00 |  327616
 Customer#000000007 | 1995-10-01 00:00:00+00 |  190890
 Customer#000000007 | 1996-06-01 00:00:00+00 | 79104.5
 Customer#000000007 | 1997-01-01 00:00:00+00 |  181378
 Customer#000000007 | 1997-02-01 00:00:00+00 |  151988
 Customer#000000007 | 1998-02-01 00:00:00+00 | 31698.3
 Customer#000000007 | 1998-07-01 00:00:00+00 |  286525
(13 rows)

dbt3=# delete from orders where o_custkey in ( select c_custkey from customer where c_name = 'Customer#000000007' );
DELETE 16

dbt3=# select * from customer_monthly_totalprice where c_name='Customer#000000007';
       c_name       |       date_trunc       |   sum
--------------------+------------------------+---------
 Customer#000000007 | 1992-03-01 00:00:00+00 |  322432
 Customer#000000007 | 1992-04-01 00:00:00+00 |  434825
 Customer#000000007 | 1993-06-01 00:00:00+00 |  501704
 Customer#000000007 | 1993-11-01 00:00:00+00 | 80777.5
 Customer#000000007 | 1994-02-01 00:00:00+00 |  176604
 Customer#000000007 | 1994-12-01 00:00:00+00 |  192318
 Customer#000000007 | 1995-09-01 00:00:00+00 |  327616
 Customer#000000007 | 1995-10-01 00:00:00+00 |  190890
 Customer#000000007 | 1996-06-01 00:00:00+00 | 79104.5
 Customer#000000007 | 1997-01-01 00:00:00+00 |  181378
 Customer#000000007 | 1997-02-01 00:00:00+00 |  151988
 Customer#000000007 | 1998-02-01 00:00:00+00 | 31698.3
 Customer#000000007 | 1998-07-01 00:00:00+00 |  286525
(13 rows)

dbt3=# refresh materialized view customer_monthly_totalprice;
REFRESH MATERIALIZED VIEW
dbt3=# select * from customer_monthly_totalprice where c_name='Customer#000000007';
 c_name | date_trunc | sum
--------+------------+-----
(0 rows)

dbt3=#

■まとめ