This product has been developed internally at habitissimo for allowing developers to get the database instances they need for development as fast as possible.
This project consists on a service which will import a collection of databases periodically. This databases become templates for the final users.
An user can ask for a database instance from any template available and have it fully functional and loaded with data within seconds, no matter how big is the database, this databases can be destroyed at any moment to request a new instance with fresh data.
The main concern we where having in our development process was importing database backups in our development instances, loading this backups by tradicional means (importing a mysqldump file) could take almost an hour, we could use other metohds like innobackupex, but this would mean developers had to download huge files (even with compression) trading speed in import time by slownes in download time.
This solution is being used to provide a variety of databases ranging from a few megabytes up to several gigabytes, all of them are provisioned within seconds (something between 3 or 5 seconds).
You put your sql backups in a folder and run the updater command, this will import the databases and prepare them as templates. This is the slow part, we run it at nights so developers can have acces to yesterday's data in the morning.
The backups are loaded into a dockerized mysql instance, this docker container binds the datadir to a host volume stored on a filesystem with Copy On Write support.
Once the templates have been loaded the script stops the template database instances.
Every time a user asks for a new database the service performs a copy on write from the template to a new directory, this directory is mounted as a volume for a new mysql docker instance launched for this user. As the operation is performed against a COW filesystem the operation is both fast and space efficient.
Finally the service responds with access data required to use the database.
- Use docker volume API instead of hacking arround with volume bindings
- Create adapters for postgresql and mongodb
- Update testsuite, broken after refactoring
MyAAS has been designed with mysql in mind, but the implementation is database agnostic and can be adapted easily to work with any type of database which stores data in disk.
Look for MysqlDatabase adapter to have an idea of how easy is to support new databases, you just need to extend AbstractDatabase and define a few properties.
If you have problems using this service open an issue.