一款PostgreSQL WAL日志解析工具: wal2json

Posted 2021-04-30 PostgreSQL中文社区

译者简介

钟硕， 现供职于迪卡侬，PostgreSQL & Oracle DBA

Debezium 属于红帽开源项目， wal2json作为其组成部分可以提供基于PG库级别的DML日志挖掘工作。

Debezium针对几类主流数据库的连接器：

https://debezium.io/docs/connectors/

Wal2json安装页面：

https://debezium.io/docs/install/postgres-plugins/

Debezium 目地在于提供一个分布式的平台，将数据库日志中的事件记录转化为事件流，使得外部应用能够对数据库中的行级操作做出快速响应。Debezium 可以建立在Apache kafka的上层，为kafka connect提供可兼容的连接器用于监控和管理特定的数据库。

wal2json工具的安装

自从PostgreSQL 9.4 支持logical级别的WAL后，PostgreSQL可以通过逻辑解码的方式（基于行或语句的logical replication）来解读WAL中的内容。

Logical Decoding Output Plugins

https://www.postgresql.org/docs/9.4/logicaldecoding-output-plugin.html

WAL_LEVEL支持的四个级别

https://www.postgresql.org/docs/9.4/runtime-config-wal.html

 

通过slot在从节点解码WAL中的数据变更流进行重演实现数据库的逻辑复制。

安装：

export PATH="$PATH:/usr/pgsql-9.5/bin"

注意：这里一定要确认系统环境中找到正确版本的pg_config（如果安装过多个版本的PostgreSQL）

git clone https://github.com/eulerto/wal2json -b master --single-branchcd wal2jsonmake && make install 

正确安装后的输出：

Cloning into 'wal2json'...remote: Enumerating objects: 47, done.remote: Counting objects: 100% (47/47), done.remote: Compressing objects: 100% (24/24), done.remote: Total 496 (delta 26), reused 38 (delta 23), pack-reused 449Receiving objects: 100% (496/496), 204.39 KiB | 0 bytes/s, done.Resolving deltas: 100% (345/345), done.
gcc -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Wendif-labels -Wmissing-format-attribute -Wformat-security -fno-strict-aliasing -fwrapv -fexcess-precision=standard -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -fPIC -I. -I./ -I/usr/pgsql-9.5/include/server -I/usr/pgsql-9.5/include/internal -D_GNU_SOURCE -I/usr/include/libxml2  -I/usr/include  -c -o wal2json.o wal2json.cgcc -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Wendif-labels -Wmissing-format-attribute -Wformat-security -fno-strict-aliasing -fwrapv -fexcess-precision=standard -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector-strong --param=ssp-buffer-size=4 -grecord-gcc-switches -m64 -mtune=generic -fPIC -L/usr/pgsql-9.5/lib -Wl,--as-needed  -L/usr/lib64 -Wl,--as-needed -Wl,-rpath,'/usr/pgsql-9.5/lib',--enable-new-dtags  -shared -o wal2json.so wal2json.o
/usr/bin/mkdir -p '/usr/pgsql-9.5/lib'/usr/bin/install -c -m 755  wal2json.so '/usr/pgsql-9.5/lib/'

PostgreSQL服务端配置

shared_preload_libraries = 'wal2json'wal_level = logicalmax_wal_senders = 4 max_replication_slots = 4

创建具有Replication和Login授权的用户

CREATE ROLE <name> REPLICATION LOGIN;

修改pg_hba.conf，使该用户可以远程或本地访问数据库

############ REPLICATION ##############local   replication     <name>                              trusthost    replication     <name>    127.0.0.1/32     trust host    replication     <name>    ::1/128              trust

建立测试环境

CREATE DATABASE test;
CREATE TABLE test_table (    id char(10) NOT NULL,    code        char(10),    PRIMARY KEY (id));

另开一个会话窗口

pg_recvlogical -d test --slot test_slot --create-slot -P wal2jsonpg_recvlogical -d test --slot test_slot --start -o pretty-print=1 -f -

进行一些基本的DML操作

test=# INSERT INTO test_table (id, code) VALUES('id1', 'code1');INSERT 0 1test=# update test_table set code='code2' where id='id1';UPDATE 1test=# delete from test_table where id='id1';DELETE 1

对应DML的输出：

INSERT

{        "change": [                {                        "kind": "insert",                        "schema": "mdmv2",                        "table": "test_table",                        "columnnames": ["id", "code"],                        "columntypes": ["character(10)", "character(10)"],                        "columnvalues": ["id1       ", "code1     "]                }        ]}

UPDATE

{        "change": [                {                        "kind": "update",                        "schema": "mdmv2",                        "table": "test_table",                        "columnnames": ["id", "code"],                        "columntypes": ["character(10)", "character(10)"],                        "columnvalues": ["id1       ", "code2     "],                        "oldkeys": {                                "keynames": ["id"],                                "keytypes": ["character(10)"],                                "keyvalues": ["id1       "]                        }                }        ]}

DELETE

{        "change": [                {                        "kind": "delete",                        "schema": "mdmv2",                        "table": "test_table",                        "oldkeys": {                                "keynames": ["id"],                                "keytypes": ["character(10)"],                                "keyvalues": ["id1       "]                        }                }        ]}

也可以用PostgreSQL提供的REPLICA IDENTITY来决定对表UPDATE和DELETE操作时，logical日志输出信息的详细程度。

DEFAULT  逻辑日志中包含表中主键列被UPDATE或DELETE的前值的信息。

NOTHING 逻辑日志中不包含表中任何UPDATE或DELETE变更的信息。

FULL 逻辑日志中包含表中被UPDATE或DELETE列的前值的整行信息。

USING INDEX <index_name>仅包含指定索引中所有列的前值信息。

ALTER TABLE test_table REPLICA IDENTITY USING INDEX test_table_pkey；

{        "change": [                {                        "kind": "update",                        "schema": "mdmv2",                        "table": "test_table",                        "columnnames": ["id", "code"],                        "columntypes": ["character(10)", "character(10)"],                        "columnvalues": ["id1       ", "code2     "],                        "oldkeys": {                                "keynames": ["id"],                                "keytypes": ["character(10)"],                                "keyvalues": ["id1       "]                        }                }        ]}

注意：    

 wal2json 插件无法很好的处理引号标识符    

wal2json 插件无法输出没有主键的表的事件信息    

wal2json 插件不支持特殊浮点类型的值（Nan 或 infinity）

 

参考

https://debezium.io/docs/install/postgres-plugins/

PostgreSQL中文社区欢迎广大技术人员投稿

投稿邮箱：[e-mail]