baskard / briskclusterami

Code for creating a self-launching Brisk Cluster in EC2.

Summary

DataStax's Amazon Machine Image is the quickest way to get a Brisk cluster up and running on EC2.

Search for AMIs by using the term:

datastax_brisk_cluster_ami

Quickstart

Launch the number of instances desired in your cluster with the User Data field set to

-s <number of instances being started>

Options

--clustersize <size> (or -s <size>)
  Used by the configuration script to evenly space the Brisk node
  tokens so each machine gets an equal share of data.  If omitted,
  nodes will receive random tokens.

--vanillanodes <size> (or -v <size>)
  Used by the configuration script to assign the first N nodes to be
  non-task tracker Brisk nodes.

--cfsreplication <size> (or -c <size>)
  Sets the cfsreplication factor at startup.

--opscenter <user>:<pass> (or -o <user>:<pass>)
  Installs OpsCenter using the provided username and password recieved
  during OpsCenter registration.
  
  Visit http://www.datastax.com/opscenter for a free registration.

--paidopscenter <user>:<pass> (or -p <user>:<pass>)
  Installs OpsCenter using the provided username and password recieved
  during OpsCenter registration for paying customers.

--clustername <name> (or -n <name>)
  Assigns the cluster with a chosen cluster name.

Step-by-step

Visit http://www.datastax.com/docs/0.8/brisk/install_brisk_ami for full installation instructions.

Post-install

To stop Brisk, simply run

ps auwx | grep brisk

and kill the associated PID without

sudo kill <PID>

To start Brisk again, simply run

sudo ~/brisk/bin/brisk cassandra -t

for a Brisk node, or

sudo ~/brisk/bin/brisk cassandra

for a vanilla Cassandra node.

Important note on VM restarts

EC2 is an unusual environment because a VM will get a different IP if it reboots. If you reboot all of your machines, they will contact the reflector again and reconfigure themselves to be able to talk to each other again, but if you reboot a single one, it won't be able to rejoin the others without manual intervention (set the seed in cassandra.yaml to one of the other nodes).

Adding nodes to the cluster

Adding nodes must currently be done manually (that is, you can create extra nodes with the same clustername, but you'll have to edit token, bootstrap setting, and seed manually to get them to join the existing cluster). See http://wiki.apache.org/cassandra/Operations for details.

Implementation details

See FILES.txt for a description of how the scripts here configure the AMI for Brisk.

Branching details

Feel free to fork off this project and offer any suggestions that you find along the way.

Also, if you're interested in the whole process: read up on the saving process here: http://www.datastax.com/dev/blog/personalizing-your-own-brisk-ami