Skip to content

Latest commit

 

History

History
59 lines (44 loc) · 2.26 KB

README.md

File metadata and controls

59 lines (44 loc) · 2.26 KB
Raw

Deploy multiple slaves to your MySQL database on dotCloud

This is an experimental recipe to have more than 1 slave on your MySQL database on dotCloud.

It works like this:

  1. Deploy one MySQL master/slave service (as usual).
  2. Add a bunch of separate MySQL standalone (non scaled) services.
  3. Reconfigure those separate services as slaves of the first one.
  4. Manually tweak your code to connect either to the master/slave service (for write ops) or to one of the standalone slaves (for read-only things).

This is alpha, and can probably break in many ways. If the main master/slave service does a failover, you might have to manually resync the slaves. The procedure is not complex, but it can take some time if you have a big database.

How to use it

Add the content of the provided dotcloud.yml file to your dotcloud.yml build file. You don't have to include the dbwrite section if you want to use an existing database. You can add more dbread services if you need to.

Add the dbmanager directory to your code.

If you changed the name of the dbwrite service, or if you added more slaves, you need to edit dbmanager/startsync to set MASTER and SLAVES environment variables.

Push.

Exec dotcloud run myapp.dbmanager ./startsync.

Important note about credentials

After runing the startsync script, the slave DB credentials will be wiped out, and replaced with those of the master DB. So, when connecting to the slave DB dbread1, remember to use the host and port shown by dotcloud info myapp.dbread1, but the login and password shown by dotcloud info myapp.dbwrite!

Just to make it clear: dotcloud info myapp.dbread1 will still show the old, invalid credentials for the slave databases. Don't believe dotcloud info in that case: it does not know that you have changed the MySQL root password (by setting up synchronization with another database).

Todo

  • guess dbwrite and dbread services by parsing environment.json
  • fix startsync so it waits for all servers to be ready
  • run startsync automatically
  • write a crude HTTP service to expose replication status
  • add a long running process to check replication status, and optionally force a full resync
  • add a mysql proxy service to automate query routing