a1exsh / cassandra-etcd-seed-provider

A Cassandra SeedProvider that fetches the list of seed nodes from Etcd

Cassandra Seed Providers

Cassandra gets its list of seed nodes from an implementation of the SeedProvider interface. The most common implementation is bundled with Cassandra, the SimpleSeedProvider. The SimpleSeedProvider reads the list of seed nodes from the cassandra.yaml configuration file.

If you're deploying Cassandra as an immutable stack, changing the list of seed nodes at runtime can be challenging. If you need to deploy it on your AWS VPC using only dynamic IP addresses it becomes even more complicated.

If you find yourself in any of these circumstances this project may help you.

Etcd Seed Provider

The EtcdSeedProvider addresses those challenges by fetching the list of seed nodes from an external Etcd cluster which can be dynamically updated.

It's up to the process that maintains the list of seed nodes to delete dead entries and replace them properly by new ones. Such process should store the entries in a specific format the EtcdSeedProvider understands.

The seeds nodes keys should be stored under the url parameter setting. The key name itself it not relevant and the value should be the IP address or hostname.

Etcd format

Assuming your cluster configuration root is under http://my-etcd-cluster.domain.com/v2/keys/cassandra/my-cluster-name/seeds the expected format would be:

| Key                                                          | Value     |
|--------------------------------------------------------------|-----------|
| /v2/keys/cassandra/my-cluster-name/seeds/meaninglesskeyname1 | 127.0.1.1 |
| /v2/keys/cassandra/my-cluster-name/seeds/meaninglesskeyname2 | 127.0.1.2 |
| /v2/keys/cassandra/my-cluster-name/seeds/meaninglesskeyname3 | 127.0.1.3 |

The above content on your Etcd would use 127.0.1.1, 127.0.1.2 and 127.0.1.3 as the seed nodes for your cluster.

The key names holding the ip address or hostname for each seed node is effectively meaningless. The EtcdSeedProvider just gets the values for all the child nodes under the url setting.

Howto

First you have to copy the latest jar file for this library to your cassandra/lib sub-folder to make sure the EtcdSeedProvider is available. You can build it yourself or check the Download section below.

You then have to change the seed provider class_name configuration and set the url parameter in your cassandra.yaml configuration file.

For example:

seed_provider:
    - class_name: org.zalando.cassandra.locator.EtcdSeedProvider
    parameters:
        - url: http://my-etcd-cluster.domain.com/v2/keys/cassandra/my-cluster-name/seeds

Download

You can download the latest released version from bintray

Known limitations and To Do list

• Only the IP addresses are stored, without any further metadata like AZ