## Cloud Native Deployments of Cassandra using Kubernetes
The following document describes the development of a _cloud native_ [Cassandra](http://cassandra.apache.org/) deployment on Kubernetes. When we say _cloud native_ we mean an application which understands that it is running within a cluster manager, and uses this cluster management infrastructure to help implement the application. In particular, in this instance, a custom Cassandra ```SeedProvider``` is used to enable Cassandra to dynamically discover new Cassandra nodes as they join the cluster.
This example assumes that you have a Kubernetes cluster installed and running, and that you have installed the ```kubectl``` command line tool somewhere in your path. Please see the [getting started](https://github.com/GoogleCloudPlatform/kubernetes/tree/master/docs/getting-started-guides) for installation instructions for your platform.
This example also has a few code and configuration files needed. To avoid typing these out, you can ```git clone``` the Kubernetes repository to you local computer.
In Kubernetes, the atomic unit of an application is a [_Pod_](../../docs/user-guide/pods.md). A Pod is one or more containers that _must_ be scheduled onto the same host. All containers in a pod share a network namespace, and may optionally share mounted volumes.
There are a few things to note in this description. First is that we are running the ```kubernetes/cassandra``` image. This is a standard Cassandra installation on top of Debian. However it also adds a custom [```SeedProvider```](https://svn.apache.org/repos/asf/cassandra/trunk/src/java/org/apache/cassandra/locator/SeedProvider.java) to Cassandra. In Cassandra, a ```SeedProvider``` bootstraps the gossip protocol that Cassandra uses to find other nodes. The ```KubernetesSeedProvider``` discovers the Kubernetes API Server using the built in Kubernetes discovery service, and then uses the Kubernetes API to find new nodes (more on this later)
You may also note that we are setting some Cassandra parameters (```MAX_HEAP_SIZE``` and ```HEAP_NEWSIZE```) and adding information about the [namespace](https://github.com/GoogleCloudPlatform/kubernetes/blob/master/docs/namespaces.md). We also tell Kubernetes that the container exposes both the ```CQL``` and ```Thrift``` API ports. Finally, we tell the cluster manager that we need 0.5 cpu (0.5 core).
In theory could create a single Cassandra pod right now but since `KubernetesSeedProvider` needs to learn what nodes are in the Cassandra deployment we need to create a service first.
In Kubernetes a _[Service](../../docs/user-guide/services.md)_ describes a set of Pods that perform the same task. For example, the set of Pods in a Cassandra cluster can be a Kubernetes Service, or even just the single Pod we created above. An important use for a Service is to create a load balancer which distributes traffic across members of the set of Pods. But a _Service_ can also be used as a standing query which makes a dynamically changing set of Pods (or the single Pod we've already created) available via the Kubernetes API. This is the way that we use initially use Services with Cassandra.
The important thing to note here is the ```selector```. It is a query over labels, that identifies the set of _Pods_ contained by the _Service_. In this case the selector is ```name=cassandra```. If you look back at the Pod specification above, you'll see that the pod has the corresponding label, so it will be selected for membership in this Service.
Of course, a single node cluster isn't particularly interesting. The real power of Kubernetes and Cassandra lies in easily building a replicated, scalable Cassandra cluster.
In Kubernetes a _[Replication Controller](../../docs/user-guide/replication-controller.md)_ is responsible for replicating sets of identical pods. Like a _Service_ it has a selector query which identifies the members of it's set. Unlike a _Service_ it also has a desired number of replicas, and it will create or delete _Pods_ to ensure that the number of _Pods_ matches up with it's desired state.
Replication controllers will "adopt" existing pods that match their selector query, so let's create a replication controller with a single replica to adopt our existing Cassandra pod.
Most of this replication controller definition is identical to the Cassandra pod definition above, it simply gives the resplication controller a recipe to use when it creates new Cassandra pods. The other differentiating parts are the ```selector``` attribute which contains the controller's selector query, and the ```replicas``` attribute which specifies the desired number of replicas, in this case 1.
Notice that one of the pods has the human readable name ```cassandra``` that you specified in your config before, and one has a random string, since it was named by the replication controller.
To prove that this all works, you can use the ```nodetool``` command to examine the status of the cluster. To do this, use the ```kubectl exec``` command to run ```nodetool``` in one of your Cassandra pods.
