pg_chameleon MySQL to PostgreSQL replica made easy

pg chameleon
MySQL to PostgreSQL replica made easy
Federico Campoli
Transferwise
PGCon, Ottawa
01 Jun 2018
http://www.pgdba.org
@4thdoctor scarf
Federico Campoli (Transferwise) pg chameleon PGCon, Ottawa01 Jun 2018 1 / 46

Few words about the speaker
Born in 1972
Passionate about IT since 1982
mostly because of the TRON movie

Born in 1972
Joined the Oracle DBA secret society in 2004
In love with PostgreSQL since 2006

Born in 1972
Joined the Oracle DBA secret society in 2004
In love with PostgreSQL since 2006
Devrim PostgreSQL tattoo’s copycat
Works at Transferwise as Data Engineer

Disclaimer
I’m not a developer
I’m a DBA...

Disclaimer
I’m a DBA...which means being hated by everybody and hating everybody

Disclaimer
So, to put things in the right perspective...

Disclaimer
So, to put things in the right perspective...I use tabs

Palpatine

Table of contents
1 History
2 MySQL Replica in a nutshell
3 A chameleon in the middle
4 Replica in action
5 Lessons learned
6 Wrap up

History

The beginnings
Years 2006/2012
neo my2pg.py
I wrote the script because of a struggling phpbb on MySQL
The database migration was successful
However phpbb didn’t work very well with PostgreSQL.1
1Opening a new connection for each query is not the smartest thing to do.

The beginnings
Years 2006/2012
neo my2pg.py
The script is written in python 2.6
It’s a monolith script
And it’s slow, very slow

The beginnings
Years 2006/2012
neo my2pg.py
The script is written in python 2.6
It’s a monolith script
And it’s slow, very slow
It’s a good checklist for things to avoid when coding
https://github.com/the4thdoctor/neo my2pg

I’m not scared of using the ORMs
Years 2013/2015
First attempt of pg chameleon
Developed in Python 2.7
Used SQLAlchemy for extracting the MySQL’s metadata
Proof of concept only
It was built during the years of the life on a roller coaster2
Therefore it was a just a way to discharge frustration
2Recording available here: http://www.pgbrighton.uk/post/backup recovery/

I’m not scared of using the ORMs
Years 2013/2015
First attempt of pg chameleon
Developed in Python 2.7
Used SQLAlchemy for extracting the MySQL’s metadata
Proof of concept only
It was built during the years of the life on a roller coaster2
Therefore it was a just a way to discharge frustration
Abandoned after a while
SQLAlchemy’s limitations were frustrating as well (see slide 3)
And pgloader did the same job much much better
2Recording available here: http://www.pgbrighton.uk/post/backup recovery/

pg chameleon reborn
Year 2016
I needed to replicate the data data from MySQL to PostgreSQL
http://tech.transferwise.com/scaling-our-analytics-database/

pg chameleon reborn
Year 2016
I needed to replicate the data data from MySQL to PostgreSQL
http://tech.transferwise.com/scaling-our-analytics-database/
The amazing library python-mysql-replication allowed me build a proof of
concept
Evolved later in pg chameleon 1.x
Kudos to the python-mysql-replication team!
https://github.com/noplay/python-mysql-replication

pg chameleon 1.x
Developed on the London to Brighton commute
Released as stable the 7th May 2017
Followed by 8 bugﬁx releases

pg chameleon 1.x
Developed on the London to Brighton commute
Released as stable the 7th May 2017
Followed by 8 bugﬁx releases
Compatible with CPython 2.7/3.3+
No more SQLAlchemy
The MySQL driver changed from MySQLdb to PyMySQL
Command line helper
Supports type override on the ﬂy (Danger!)
Installs in virtualenv and system wide via pypi
Can detach the replica for minimal downtime migrations

pg chameleon versions 1’s limitations
All the aﬀected tables are locked in read only mode during the init replica
process
During the init replica the data is not accessible

process
The tables for being replicated require primary keys
No daemon, the process always stays in foreground
Single schema replica
One process per each schema
Network ineﬃcient

process
The tables for being replicated require primary keys
No daemon, the process always stays in foreground
Single schema replica
One process per each schema
Network ineﬃcient
Read and replay not concurrent with risk of high lag
The optional threaded mode very ineﬃcient and fragile
A single error in the replay process and the replica is broken

MySQL Replica in a nutshell

MySQL Replica
The MySQL replica is logical
When the replica is enabled the data changes are stored in the master’s
binary log ﬁles
The slave gets from the master’s binary log ﬁles
The slave saves the stream of data into local relay logs
The relay logs are replayed against the slave

MySQL Replica

Log formats
MySQL have three ways of storing the changes in the binary logs.
STATEMENT: It logs the statements which are replayed on the slave.
It’s the best solution for the bandwidth. However, when replaying statements
with not deterministic functions this format generates diﬀerent values on the
slave (e.g. using an insert with a column autogenerated by the uuid function).
ROW: It’s deterministic. This format logs the row images.
MIXED takes the best of both worlds. The master logs the statements unless
a not deterministic function is used. In that case it logs the row image.
All three formats always log the DDL query events.
The python-mysql-replication library and therefore pg chameleon, require the
ROW format to work properly.

A chameleon in the middle

pg chameleon
pg chameleon mimics a mysql slave’s behaviour
It performs the initial load for the replicated tables
It connects to the MySQL replica protocol
It stores the row images into a PostgreSQL table
A plpgSQL function decodes the rows and replay the changes
It can detach the replica for minimal downtime migrations
PostgreSQL acts as relay log and replication slave

MySQL replica + pg chameleon

pg chameleon 2.0 #1
Developed at the pgconf.eu 2017 and on the commute
Released as stable the 1st of January 2018
Compatible with python 3.3+
Installs in virtualenv and system wide via pypi
Replicates multiple schemas from a single MySQL into a target PostgreSQL
database
Conservative approach to the replica. Tables which generate errors are
automatically excluded from the replica
Daemonised replica process with two distinct subprocesses, for concurrent
read and replay

pg chameleon 2.0 #2
Soft locking replica initialisation. The tables are locked only during the copy.
Rollbar integration for a simpler error detection and messaging
Experimental support for the PostgreSQL source type
The tables are loaded in a separate schema which is swapped with the
existing.
This approach requires more space but it makes the init a replica virtually
painless, leaving the old data accessible until the init replica is complete.
The DDL are translated in the PostgreSQL dialect keeping the schema in
sync with MySQL automatically

Version 2.0’s limitations
Tables for being replicated require primary or unique keys
When detaching the replica the foreign keys are created always ON
DELETE/UPDATE RESTRICT
The source type PostgreSQL supports only the init replica process

Replica initialisation
The replica initialisation follows the same workﬂow as stated on the mysql online
manual.
Flush the tables with read lock
Get the master’s coordinates
Copy the data
Release the locks
However...

Replica initialisation
The replica initialisation follows the same workﬂow as stated on the mysql online
manual.
Flush the tables with read lock
Get the master’s coordinates
Copy the data
Release the locks
However...
pg chameleon ﬂushes the tables with read lock one by one. The lock is held only
during the copy.
The log coordinates are stored in the replica catalogue along the table’s name and
used by the replica process to determine whether the table’s binlog data should be
used or not.
The replica starts inconsistent and gains consistency over time.

Fallback on failure
The data is pulled from mysql using the CSV format in slices. This approach
prevents the memory overload.
Once the ﬁle is saved then is pushed into PostgreSQL using the COPY command.
However...

Fallback on failure
The data is pulled from mysql using the CSV format in slices. This approach
prevents the memory overload.
Once the ﬁle is saved then is pushed into PostgreSQL using the COPY command.
However...
COPY is fast but is single transaction
One failure and the entire batch is rolled back
If this happens the procedure loads the same data using the INSERT
statements
Which can be very slow
The process attempts to clean the NUL markers which are allowed by MySQL
If the row still fails on insert then it’s discarded

Replica in action

MySQL configuration
The mysql configuration file is usually stored in /etc/mysql/my.cnf
To enable the binary logging find the section [mysqld] and check that the
following parameters are set.
binlog_format= ROW
log-bin = mysql-bin
server-id = 1
binlog-row-image = FULL

MySQL user for replica
Setup a replication user on MySQL
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;
In our example we are using the sakila test database.
https://dev.mysql.com/doc/sakila/en/

PostgreSQL setup
Add an user on PostgreSQL capable to create schemas and relations in the
destination database
CREATE USER usr_replica WITH PASSWORD ’replica ’;
CREATE DATABASE db_replica WITH OWNER usr_replica;

Install pg chameleon
Install pg chameleon and create the configuration files
pip install pip --upgrade
pip install pg_chameleon
chameleon set_configuration_files
cd ~/.pg_chameleon/configuration
cp config-example.yml default.yml
Edit the file default.yml setting the correct values for connection and source.

Conﬁgure global settings in default.yaml
PostgreSQL Connection
pg conn:
host: " localhost "
p or t : " 5432 "
u s e r : " usr_replica "
password: " replica "
database: " db_replica "
c h a r s e t : " utf8 "

pg conn:
host: " localhost "
p or t : " 5432 "
Rollbar conﬁguration
r o l l b a r k e y : ’< rollbar_long_key>’
r o l l b a r e n v : ’pgcon - demo ’

pg conn:
host: " localhost "
p or t : " 5432 "
Rollbar conﬁguration
r o l l b a r k e y : ’< rollbar_long_key>’
r o l l b a r e n v : ’pgcon - demo ’
Type override (optional)
t y p e o v e r r i d e :
" tinyint (1) ":
o v e r r i d e t o : b o o l e a n
o v e r r i d e t a b l e s :
- "*"

Conﬁgure the mysql source
s o u r c e s :
mysql:
db conn:
host: " localhost "
po r t : " 3306 "
c h a r s e t : ’utf8 ’
connect timeout: 10

s o u r c e s :
mysql:
db conn:
host: " localhost "
po r t : " 3306 "
connect timeout: 10
schema mappings:
s a k i l a : l o x o d o n t a a f r i c a n a

s o u r c e s :
mysql:
db conn:
host: " localhost "
po r t : " 3306 "
connect timeout: 10
schema mappings:
s a k i l a : l o x o d o n t a a f r i c a n a
l i m i t t a b l e s :
s k i p t a b l e s :
g r a n t s e l e c t t o :
- u s r r e a d o n l y
l o c k t i m e o u t : " 120 s"
m y s e r v e r i d : 100
r e p l i c a b a t c h s i z e : 10000
rep l ay max row s: 10000
b a t c h r e t e n t i o n : ’1 day ’
copy max memory: " 300 M"
copy mode: ’file ’
o u t d i r : /tmp
s l e e p l o o p : 1
o n e r r o r r e p l a y : c o n t i n u e
o n e r r o r r e a d : c o n t i n u e
auto maintenance: "1 day "
type: mysql

Add the source and initialise the replica
Add the source mysql and initialise the replica for it. We are using debug in order
to get the logging on the console.
chameleon create_replica_schema --debug
chameleon add_source --config default --source mysql --debug
chameleon init_replica --config default --source mysql --debug

Start the replica
Start the replica process
chameleon start_replica --config default --source mysql

Start the replica
Start the replica process
chameleon start_replica --config default --source mysql
Show the replica status
chameleon show_status --config default --source mysql

Time for a demo
Demo!
The demo will fail miserably for sure and you will hate this project forever.

Lessons learned

Strictness is an illusion. MySQL doubly so
MySQL’s lack of strictness is not a mystery.
The funny way the default with NOT NULL is managed by MySQL can break the
replica.
Therefore any ﬁeld with NOT NULL added after the initialisation are created
always as NULLable in PostgreSQL.

The DDL. A real pain in the back
I initially tried to use sqlparse for tokenising the DDL emitted by MySQL.
Unfortunately didn’t worked as I expected.

So I decided to use the regular expressions.
Some people, when confronted with a problem,
think "I know, I’ll use regular expressions."
Now they have two problems.
-- Jamie Zawinski

So I decided to use the regular expressions.
Some people, when confronted with a problem,
think "I know, I’ll use regular expressions."
Now they have two problems.
-- Jamie Zawinski
MySQL even in ROW format emits the DDL as statements
The class sql token uses the regular expressions to tokenise the DDL
The tokenised data is used to build the DDL in the PostgreSQL dialect

Wrap up

To boldly go where no chameleon has gone before
Short team goals, version 2.0
Re sync automatically the tables when they error on replay
Improve the replay speed and cpu eﬃciency
GTID support for MySQL source
Medium term goals version 2.1
Parallel copy and index creation in order to speed up the init replica process
Logical replica from PostgreSQL
Improve the default column handling

Igor, the green little guy
The chameleon logo has been developed by Elena Toma, a talented Italian Lady.
https://www.facebook.com/Tonkipapperoart/
The name Igor is inspired by Martin Feldman’s Igor portraited in Young
Frankenstein movie.

Feedback please!
Please report any issue on github and follow pg chameleon on twitter for the
announcements.
https://github.com/the4thdoctor/pg chameleon
@pg chameleon

Did you say hire?
WE ARE HIRING!
https://transferwise.com/jobs/

That’s all folks!
Thank you for listening!
Any questions?
Please be very basic, I’m just an electrician after all.

Image credits
Palpatine,Dr. Evil disclaimer,It could work. Young Frankenstein source
memegenerator
MySQL Image source, WikiCommons
Hard Disk image, source WikiCommons
Tron image, source Tron Wikia
Twitter icon, source Open Icon Library
The PostgreSQL logo, copyright the PostgreSQL global development group
Boromir get rid of mysql, source imgflip
Morpheus, source imgflip
Keep calm chameleon, source imgflip
The dolphin picture - Copyright artnoose
Perseus, Framed - Copyright Federico Campoli
Pinkie Pie that’s all folks, Copyright by dan232323, used with permission
Doom, source RetroPie

License
This document is distributed under the terms of the Creative Commons
Attribution, Not Commercial, Share Alike

pg chameleon
MySQL to PostgreSQL replica made easy
Federico Campoli
Transferwise
PGCon, Ottawa
01 Jun 2018
http://www.pgdba.org
@4thdoctor scarf

pg_chameleon MySQL to PostgreSQL replica made easy

More Related Content

What's hot (20)

Similar to pg_chameleon MySQL to PostgreSQL replica made easy (20)

More from Federico Campoli (8)

Recently uploaded (20)

pg_chameleon MySQL to PostgreSQL replica made easy