ec2cluster 0.28

Tools to work with clustered applications (PostgreSQL, Redis) on EC2

Generic clustering package for EC2, suitable for redis/postgresql etc.

This package makes it easier to deploy clustered applications such as PostgreSQL and Redis on EC2 by handling generic logic, including:

* Deciding which role the instance should assume
* Creating and updating DNS records
* Preparing the instance for its role (e.g. writing a recovery.conf file for postgres)

Basic Usage:

The default ec2cluster classes assume your EC2 instances have JSON-encoded user data containing some specific attributes. The following attributes are required:

* cluster - the name of the cluster, e.g. maindb

Install ec2cluster with pip:

pip install ec2cluster

Create a configuration file:

MASTER_CNAME = 'master.%(cluster)s.example.com'
SLAVE_CNAME = 'slave.%(cluster)s.example.com'
INIT_MASTER_SCRIPT = '/path/to/some_script.py'
INIT_SLAVE_SCRIPT = '/path/to/another_script.py'

Run ec2cluster, specifying the path to the config file:

ec2cluster init # initialise the cluster service
ec2cluster promote # promote a slave to the master role

PostgreSQL cluster:

When starting a postgres read-slave, a file named recovery.conf must be written to the postgres configuration directory. A template file is used to make it easy to customise your recovery options.

Config file:

MASTER_CNAME = 'master.%(cluster)s.example.com'
SLAVE_CNAME = 'slave.%(cluster)s.example.com'
RECOVERY_TEMPLATE = '/path/to/template.conf'
RECOVERY_FILENAME = '/var/lib/postgresql/9.1/main/recovery.conf'

In the recovery template file, specify the options required for your read-slaves. Instance metadata and userdata can be used for string replacement. For example:

standby_mode = on
recovery_target_timeline = latest
pause_at_recovery_target = false
restore_command = '/usr/bin/s3cmd --config=/var/lib/postgresql/.s3cfg get s3://%(cluster)s/archive/wal/%%f %%p'
primary_conninfo = 'host=%(master_cname) port=5432 user=postgres password=secret sslmode=disable

Note the use of “%%f” - because we are using string formatting we need to escape the percentage sign in order to end up with “%f” as required by postgres.