文档章节

MySQL 到PostgreSQL 的数据迁移工具

MtrS
 MtrS
发布于 2017/05/18 08:20
字数 2356
阅读 423
收藏 0

pg_chameleon is available on pypi for download

python 包下载地址

The documentation is available on pythonhosted

文档链接:

Discussion forum available on googlegroups

google group讨论

Please submit your bug reports on GitHub.

发现问题请报告地址

Platform and versions

平台以及版本

The tool is developed using Linux Slackware 14.2.Is currently tested with python 2.7 and python 3.6. 这个工具我们是在Linux Slackware 14.2上开发,已经在Python2.7和Python2.6上测试

The tool is developed using FreeBSD as database server with 该工具的数据库测试环境是FreeBSD 服务,同时数据为:

  • MySQL: 5.6
  • PostgreSQL: 9.5

Possible applications

应用场景

  • 分析 (数据挖掘,数据仓库)

  • 迁移

  • 数据聚合, 数据源为多个MySQL 数据库

  • Analytics

  • Migrations

  • Data aggregation from multiple MySQL databases

Features

特性

  • Read the schema and data from MySQL and restore it into a target PostgreSQL schema
  • 数据MySQL 数据库的数据模式,和存储数据,在PostgreSQL 上 还原成相应的目标模式
    • Setup PostgreSQL to act as a MySQL slave
  • 将PostgreSQL扮演成一个MySQL从库
  • Basic DDL Support (CREATE/DROP/ALTER TABLE, DROP PRIMARY KEY/TRUNCATE)
  • 基本的DDL支持(CREATE/DROP/ALTER TABLE,DROP PRIMARY KEY/TRUNCATE)
  • Discards of rubbish data which is saved in the table sch_chameleon.t_discarded_rows
  • Replica from multiple MySQL schema or servers
  • Basic replica monitoring
  • Detach replica from MySQL

Requirements

安装需知

Python: Linux操作系统 , CPython2.7/3.3 (CPython是指官方发布的那个Python,非其他衍生)

Python: CPython 2.7/3.3+ on Linux

MySQL: 5.5+

PostgreSQL: 9.5+

需要安装的拓展库地址

Optionals for building documentation

文编编译需要的工具,该工具为可选项

Caveats

警告

The replica requires the tables to have a primary key. Tables without primary key are initialised during the init_replica process but the replicadoesn't update them. 这个环境需要数据表拥有主键,没有主键的数据表会在 init_replica 的过程中初始化,但是replica 不会更新

Multiple replica sources are supported. However is required a separate process for each replica. Each replica must have a unique destination schema in PostgreSQL.

支持多个复制源,但是,每一份复制需要一个单独的进程,每个复制在PostgreSQL中的映射是唯一的 The copy_max_memory is just an estimate. The average rows size is extracted from mysql's informations schema and can be outdated.If the copy process fails for memory error check the failing table's row length and the number of rows for each slice.

copy_max_memory只是一个估计,平均行大小是从MySQL的信息架构中提取出来的,并且可能已经过时了,如果复制过程由于内存错误而失败,请检查失败的表的行长度和每个slice的行数. 当复制为空时,捕获DDL或MySQL切换到另一个日志段(ROTATE EVENT),都会有这样的情况.

The batch is processed every time the replica stream is empty, when a DDL is captured or when the MySQL switches to another log segment (ROTATE EVENT).Therefore the replica_batch_size is the limit for when a write happens in PostgreSQL. The parameter controls also the size of the batch replayed by pg_engine.process_batch.

The current implementation is sequential. 目前的实现是顺序的(sequential). Read the replica -> Store the rows -> Replays the stored rows. 读取复制-> 存储行-> 重演存储的行.

2.0的版本会改进这个地方

支持Python3.0 但是MySQL的复制,需要Python3的版本在3.3以上

使用最后接受到的事件事件戳和PostgreSQL事件戳确定滞后信息.如果MySQL是只读的,延迟会增加,由于没有复制信息会产生.

detach复制进程在PostgreSQL中重演这些序列,使得PostgreSQL可以独立工作

MySQL源中的外键会被提取,但是会创建为无效的,外键在没有ON DELETE和ON UPDATE的语句中创建。第二次会长时创建外键,如果发生错误,会日志记录,由于源的配置.

The version 2.0 will improve this aspect.

Python 3 is supported but only from version 3.3 as required by mysql-replication .

The lag is determined using the last received event timestamp and the postgresql timestamp. If the mysql is read only the lag will increase becauseno replica event is coming in.

The detach replica process resets the sequences in postgres to let the database work standalone.

The foreign keys from the source MySQL schema are extracted and created initially as NOT VALID. The foreign keys are created without the ON DELETE or ON UPDATE clauses.A second run tries to validate the foreign keys. If an error occurs it gets logged out according to the source configuration.

Quick Setup

  • Create a virtual environment (e.g. python3 -m venv venv)
  • Activate the virtual environment (e.g. source venv/bin/activate)
  • Install pgchameleon with pip install pg_chameleon. If you get an error upgrade your pip first.
  • Create a user on mysql for the replica (e.g. usr_replica)
  • Grant access to usr on the replicated database (e.g. GRANT ALL ON sakila.* TO 'usr_replica';)
  • Grant RELOAD privilege to the user (e.g. GRANT RELOAD ON . to 'usr_replica';)
  • Grant REPLICATION CLIENT privilege to the user (e.g. GRANT REPLICATION CLIENT ON . to 'usr_replica';)
  • Grant REPLICATION SLAVE privilege to the user (e.g. GRANT REPLICATION SLAVE ON . to 'usr_replica';)

Configuration parameters

The system wide install is now supported correctly.

The first time chameleon.py is executed it creates a configuration directory in $HOME/.pg_chameleon.Inside the directory there are two subdirectories.

  • config is where the configuration files live. Use config-example.yaml as template for the other configuration files. Please note the logs and pid directories with relative path will no longer work. The you should either use an absolute path or provide the home alias. Again, check the config-example.yaml for an example.
  • pid is where the replica pid file is created. it can be changed in the configuration file
  • logs is where the replica logs are saved if log_dest is file. It can be changed in the configuration file

The file config-example.yaml is stored in ~/.pg_chameleon/config and should be used as template for the other configuration files.

do not use config-example.yaml directly. The tool skips this filename as the file gets overwritten when pg_chameleon is upgraded.

Is it possible to have multiple configuration files for configuring the replica from multiple source databases. It's compulsory to chose different destination schemas on postgresql.

Each source requires to be started in a separate process (e.g. a cron entry).

The configuration file is a yaml file. Each parameter controls theway the program acts.

  • my_server_id the server id for the mysql replica. must be unique among the replica cluster.
  • copy_max_memory the max amount of memory to use when copying the table in PostgreSQL. Is possible to specify the value in (k)ilobytes, (M)egabytes, (G)igabytes adding the suffix (e.g. 300M).
  • my_database mysql database to replicate. a schema with the same name will be initialised in the postgres database.
  • pg_database destination database in PostgreSQL.
  • copy_mode the allowed values are 'file' and 'direct'. With direct the copy happens on the fly. With file the table is first dumped in a csv file then reloaded in PostgreSQL.
  • hexify is a yaml list with the data types that require coversion in hex (e.g. blob, binary). The conversion happens on the copy and on the replica.
  • log_dir directory where the logs are stored.
  • log_level logging verbosity. allowed values are debug, info, warning, error.
  • log_dest log destination. stdout for debugging purposes, file for the normal activity.
  • my_charset mysql charset for the copy. Please note the replica library read is always in utf8.
  • pg_charset PostgreSQL connection's charset.
  • tables_limit yaml list with the tables to replicate. If the list is empty then the entire mysql database is replicated.
  • sleep_loop seconds between a two replica batches.
  • pause_on_reindex determines whether to pause the replica if a reindex process is found in pg_stat_activity
  • sleep_on_reindex seconds to sleep when a reindex process is found
  • reindex_app_names lists the application names to check for reindex (e.g. reindexdb). This is a workaround which required for keeping the replication user unprivileged.
  • source_name this must be unique along the list of sources. The tool detects if there's a duplicate when registering a new source
  • dest_schema this is also a unique value. once the source is registered the dest_schema can't be changed anymore
  • log_days_keep: specifies the amount how many days to keep the logs which are rotated automatically on a daily basis
  • batch_retention the max retention for the replayed batches rows in t_replica_batch. The field accepts any valid interval accepted by PostgreSQL
  • out_dir the directory where the csv files are dumped during the init_replica process if the copy mode is file

Reindex detection example setup

#Pause the replica for the given amount of seconds if a reindex process is found
pause_on_reindex: Yes
sleep_on_reindex: 30

#list the application names which are supposed to reindex the database
reindex_app_names:
- 'reindexdb'
- 'my_custom_reindex'

MySQL connection parameters

mysql_conn:
    host: localhost
    port: 3306
    user: replication_username
    passwd: never_commit_passwords

PostgreSQL connection parameters

pg_conn:
    host: localhost
    port: 5432
    user: replication_username
    password: never_commit_passwords

Usage

使用

The script chameleon.py requires one of the following commands.

  • drop_schema Drops the service schema sch_chameleon with cascade option.
  • create_schema Create the service schema sch_chameleon.
  • upgrade_schema Upgrade an existing schema sch_chameleon to an newer version.
  • init_replica Create the table structure from the mysql into a PostgreSQL schema with the same mysql's database name. The mysql tables are locked in read only mode and the data is copied into the PostgreSQL database. The master's coordinates are stored in the PostgreSQL service schema. The command drops and recreate the service schema.
  • start_replica Starts the replication from mysql to PostgreSQL using the master data stored in sch_chameleon.t_replica_batch. The master's position is updated time a new batch is processed. The command upgrade the service schema if required.
  • list_config List the available configurations and their status ('ready', 'initialising','initialised','stopped','running')
  • add_source register a new configuration file as source
  • drop_source remove the configuration from the registered sources
  • stop_replica ends the replica process gracefully
  • disable_replica ends the replica process and disable the restart
  • enable_replica enable the replica process
  • sync_replica sync the data between mysql and postgresql without dropping the tables
  • show_status displays the replication status for each source, with the lag in seconds and the last received event
  • detach_replica stops the replica stream, discards the replica setup and resets the sequences in PostgreSQL to work as a standalone db.

the optional command --config followed by the configuration file name, without the yaml suffix, allow to specify different configurations.If omitted the configuration defaults to default.

Example

Create a virtualenv and activate it

python3 -m venv venv
source venv/bin/activate

Install pg_chameleon

pip install pg_chameleon

Run the script in order to create the configuration directory.

chameleon.py

cd in ~/.pg_chameleon/config and copy the configuration-example.yaml to default.yaml. Please note this is the default configuration and can be omitted when executing the chameleon.py script.

In MySQL create a user for the replica.

CREATE USER usr_replica ;
SET PASSWORD FOR usr_replica=PASSWORD('replica');
GRANT ALL ON sakila.* TO 'usr_replica';
GRANT RELOAD ON *.* to 'usr_replica';
GRANT REPLICATION CLIENT ON *.* to 'usr_replica';
GRANT REPLICATION SLAVE ON *.* to 'usr_replica';
FLUSH PRIVILEGES;

Add the configuration for the replica to my.cnf (requires mysql restart)

binlog_format= ROW
binlog_row_image=FULL
log-bin = mysql-bin
server-id = 1

If you are using a cascading replica configuration ensure the parameter log_slave_updates is set to ON.

log_slave_updates= ON

In PostgreSQL create a user for the replica and a database owned by the user

CREATE USER usr_replica WITH PASSWORD 'replica';
CREATE DATABASE db_replica WITH OWNER usr_replica;

Check you can connect to both databases from the replication system.

For MySQL

mysql -p -h derpy -u usr_replica sakila
Enter password:
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 116
Server version: 5.6.30-log Source distribution

Copyright (c) 2000, 2016, Oracle and/or its affiliates. All rights reserved.

Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective
owners.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql>

For PostgreSQL

psql  -h derpy -U usr_replica db_replica
Password for user usr_replica:
psql (9.5.5)
Type "help" for help.
db_replica=>

Setup the connection parameters in default.yaml

---
#global settings
my_server_id: 100
replica_batch_size: 1000
my_database:  sakila
pg_database: db_replica

#mysql connection's charset.
my_charset: 'utf8'
pg_charset: 'utf8'

#include tables only
tables_limit:

#mysql slave setup
mysql_conn:
    host: derpy
    port: 3306
    user: usr_replica
    passwd: replica

#postgres connection
pg_conn:
    host: derpy
    port: 5432
    user: usr_replica
    password: replica

Initialise the schema and the replica with

chameleon.py create_schema
chameleon.py add_source --config default
chameleon.py init_replica --config default

Start the replica with

chameleon.py start_replica --config default

Detaching the replica from MySQL

chameleon.py detach_replica --config default

© 著作权归作者所有

上一篇: 组合优化
下一篇: Link
MtrS
粉丝 35
博文 658
码字总数 407423
作品 0
榆林
私信 提问
使用达梦ETL数据交换平台实现mysql数据库向PostgreSQL数据库的自动化迁移

最近在做项目时,有需求需要用到postgresql数据库,而原本项目用的是mysql数据库,由于刚刚参加工作,这方面的经验就比较少,所以只得在网上查阅了各种资料。网上大多介绍的是直接手动迁移的...

Jialy
2014/01/22
813
0
An Overview of PG & MySQL Cross Replication

本博客的目的在于简述MySQL和PostgreSQL之间如何跨数据库进行复制。涉及跨数据库复制的databases一般被称作异构databases。这是将数据从一种RDBMS server复制到另一种server的一种很好的方法...

yzs的专栏
07/07
0
0
PostgreSQL 数据迁移 (从Mysql到Postgres)

前言 之前使用的数据库是MySql,现在要改成PostgreSQL,因此需要将MySql表中数据也迁移过来. 下面是一些具体操作步骤. @Author duangr @Website http://my.oschina.net/duangr/blog/182216 1....

一只小逛
2013/12/06
6.4K
1
PostgreSQL 与 MySQL 相比,优势何在?

一、 PostgreSQL 的稳定性极强, Innodb 等引擎在崩溃、断电之类的灾难场景下抗打击能力有了长足进步,然而很多 MySQL 用户都遇到过Server级的数据库丢失的场景——mysql系统库是MyISAM的,相...

leycau
2015/06/29
345
0
MySQL 不落地迁移、导入 PostgreSQL - 推荐 rds_dbsync

标签 PostgreSQL , MySQL , rdsdbsync , mysql , mysqldump , copy , mysqlfdw 背景 将MySQL数据不落地的方式导入PostgreSQL。 1 rds_dbsync (推荐使用) 《MySQL准实时同步到PostgreSQL, Gre......

德哥
04/14
0
0

没有更多内容

加载失败,请刷新页面

加载更多

oracle ORA-39700: database must be opened with UPGRADE option

ORA-01092: ORACLE instance terminated. Disconnection forced ORA-00704: bootstrap process failure ORA-39700: database must be opened with UPGRADE option 进程 ID: 3650 会话 ID: 29......

Tank_shu
今天
2
0
分布式协调服务zookeeper

ps.本文为《从Paxos到Zookeeper 分布式一致性原理与实践》笔记之一 ZooKeeper ZooKeeper曾是Apache Hadoop的一个子项目,是一个典型的分布式数据一致性的解决方案,分布式应用程序可以基于它...

ls_cherish
今天
4
0
聊聊DubboDefaultPropertiesEnvironmentPostProcessor

序 本文主要研究一下DubboDefaultPropertiesEnvironmentPostProcessor DubboDefaultPropertiesEnvironmentPostProcessor dubbo-spring-boot-project-2.7.3/dubbo-spring-boot-compatible/au......

go4it
昨天
2
0
redis 学习2

网站 启动 服务端 启动redis 服务端 在redis 安装目录下 src 里面 ./redis-server & 可以指定 配置文件或者端口 客户端 在 redis 的安装目录里面的 src 里面 ./redis-cli 可以指定 指定 连接...

之渊
昨天
2
0
Spring boot 静态资源访问

0. 两个配置 spring.mvc.static-path-patternspring.resources.static-locations 1. application中需要先行的两个配置项 1.1 spring.mvc.static-path-pattern 这个配置项是告诉springboo......

moon888
昨天
4
0

没有更多内容

加载失败,请刷新页面

加载更多

返回顶部
顶部