A tool for creating large database clusters quickly. The goal is to allow an organisation to build realistically-sized distributed database clusters in a matter of hours instead of days. Once the cluster is built, you should be able to evaluate the cluster properly by running benchmarks on it and having proper instrumentation to be able to reason logically about the performance of the cluster.
- Allocate cluster nodes, ensure passwordless SSH user with sudo is available to your Management machine.
- Decide your distribution, cluster configuration software, and database architecture.
- Enter [ConfigManager]/[Distribution] directory, edit INI and YAML files
(this will take the longest time). Some available choices are:
- Ask PalominoClusterTool to generate your cluster.
- Join the mailing list (https://groups.google.com/d/forum/palomino-cluster-tool) and IRC channels to report errors or concerns, and to discuss with other professionals.
At its core, the Palomino Cluster Tool is simply a set of playbooks, recipes, and manifests for setting up complex database clusters. There are some simple scripts to help it coordinate with your cluster management tool.
One goal is to make sure that once your cluster is set up, there is trending and alerting set up for the cluster. It is difficult to evaluate the performance of a database unless it is instrumented properly.
The project is by necessity somewhat large and complex. There are three major configuration management systems, three major Linux OS distributions, and three very desireable cluster layouts. That's 27 different sets of complex scripts that need to be generated. If your desired OS/CMS/ClusterType isn't represented, please join the mailing list and/or IRC to discuss it. There may be others with the same needs, or who are already working on it.
There is a set of slides on the various architectures to be supported, hopeful timelines, and project architecture here: http://dev.palominodb.com/time_public_html/PalominoDBClusterTool/.
You need a cluster management tool: Ansible, Chef, or Puppet for example. The end goal of the Palomino Cluster Tool remains the same, you should be able to set up the distributed database in hours, so if you do not have one of these tools set up, we recommend beginning with Ansible since it is the simplest.
Help is solicited! If you've written Chef Recipes or Puppet Manifests for setting up clusters of MySQL, HBase, Cassandra, MongoDB, or other distributed database systems, please contact us! We'd love to hear about it and welcome your participation in the project.
- Tim Ellis, CTO & Principal Architect, Palomino - Project Leader
- René Cannaò, Senior DBA, Data Architect, Palomino
- Charlie Killian, Director of Engineering, Palomino
- Cliff Erson, Riot Games
- Jamie Winsor, Riot Games
- Istvan Szukacs, Riot Games
The project and all related scripts, playbooks, manifests, recipes are Copyrighted works by the respective authors (noted in the source files themselves) and distributed under the terms of the Apache License, version 2.0.