How To Create a Cassandra Cluster in AWS Part 2

Now that I have these three machines, how do I log onto them and start to use them.

From the EC2 Dashboard, select the Instances tab on the left side. This will give you an overview of all the instances you have running.

You can select any of the instances from the grid and view all the properties about that instance. The most important detail for us here is the public IP address of these machines. Once you have that, you can SSH into the machine.

From your local machine, you’ll need to open a terminal. I’m currently running on a Mac, so my terminal will do. If you’re on a Windows machine, you’ll need something like PuTTy or cmder. I’ve use both, but prefer the latter.

From your terminal, run the following command. Note, the path to your pem file and your IP will differ.

Using the public IP of your three instances, you can access them via variations of the above command.

There are three ways to install Cassandra. The first is via a tarball that you download, unpack, and run. The second is as a package that will install itself as a service on the Linux box. The third is to download the source code and compile it. I will cover the first two methods, as I believe they are the most common. If you can compile source code, then you’re probably above this tutorial.

Before we can start, we need to make sure that the machines have the latest version of Java 7 or 8 installed.

Conveniently, my machines don’t have any java installed, so I’ll get to download and install them. My preference is to use the full Oracle JDK 8 for Cassandra.

Method 1: Install Cassandra from binary We are looking to install the tarball of Cassandra from the Apache website.

Next we un-compress the tarball.

Change directory into the newly created cassandra folder.

Start the cassandra process in the foreground.

Now you see the log output to the screen of what is going on while Cassandra starts up. Soon it will stop printing output and have line indicating that it is “listening for thrift clients” or “listening for cql clients” (depending on version). This is your indicator that Cassandra is now ready to take reads/writes.

At this point, we’ve got a single instance of Cassandra running in the foreground. Now open another terminal window and log into the same IP that you just started Cassandra on.

This will start the interactive Cassandra Query Language SHell. Viola, you officially have Cassandra running from a tarball. Now to try this again as a service.

From the terminal window that you had the cassandra process running in the foreground, press Control-C to kill the foreground process. If you happen to have lost that window, you can run the following to get the pid that Cassandra is running under and can kill it.

This second method installs Cassandra as a service on the linux machine. For this we will need a debian packaged version of Cassandra. We can either get this from the Apache Dist, or from Datastax. I like to use DataStax’s repo.

If you were to update apt-get now, you’d get an error that you don’t have the public key for the repo you just referenced. So we need to add DataStax’s public repo key.

Now we can update apt-get

Now the actual install of Cassandra 3.0

Once that is done, Cassandra will be installed as a service and will already be running. Check the status of the Cassandra service

And now we can open up the CQLSH to begin querying.

Or run run nodetool status to see what the cluster looks like.

From here, we see that we only have a single node in our cluster that is named Test Cluster running in datacenter1.

So let’s repeat the steps to install Cassandra as a service on the other two instances. Once we have that done. Let’s run nodetool status again.

None of them are talking, let’s make the changes necessary to get them communicating.

On each of the instances, open the cassandra.yaml file for editing.

We need to make the following changes to the yaml. Update the cluster_name to what you want your cluster to be called. Update the seeds with 2 of your public IPs, use these 2 for every yaml file. Update the broadcast_address with the public IP of each machine. Blank out the listen_address.

Now that we have the yaml file configured. You’ll need to stop the cassandra service, clear out the system keyspace data files, and then start the service up.

I like to tail the Cassandra log to verify that everything is happening as expected. When you receive a line with “listening for thrift clients” or “listening for cql clients” (depending on version), you know it’s back up and ready for action.

Once you have all of this done on all of your instances, you can re-run the nodetool status to get a feel for what your new cluster looks like.

Notice that you now have three entries in the output and that they are listed by their Public IPs now.

Try running cqlsh again and notice that the name of the cluster has also been updated from “Test Cluster” to "Demo Cluster”.

This tutorial has walked you through the creation of three Amazon EC2 instances. It has shown you two different ways to install Cassandra. Also, the key configuration changes and ports to be opened have been pointed out. You should be able to follow all of this information provided above and create your own Cassandra cluster in AWS.