The next interview we held at the DevNation Hackathon in San Francisco is with James Shubin.
Who are you?
I’m James Shubin, a configuration management architect in the systems engineering group at Red Hat. Before that, I was a “random hacker” who started using GlusterFS before the acquisition by Red Hat. I also write a technical blog named The Technical Blog of James.
Are you the kind of random hacker who builds things?
Yes, I do a lot of configuration management stuff. I created a Puppet module to deploy GlusterFS. I built it because I wanted to automate the testing of different configurations. When GlusterFS added a command line management utility, I had to rewrite the module from scratch, and it kept evolving as people requested features. Then a year ago, I met John Mark Walker, which lead to my work as a consultant on Gluster. After that, I was asked to stay on permanently.
Tell us about Puppet-Gluster
Puppet-Gluster is an easy way to automate or try out GlusterFS. So if you’re testing GlusterFS, evaluating it, trying out certain setups or parameters for performance, or testing and managing a production set up…it can do all those things. From my laptop I can build a whole Gluster pool in 10 to 15 minutes.
Puppet-Gluster is a Puppet module for the Puppet configuration management tool. For example, if you have a RAID array to partition--a technical task--you have to ensure the 512-byte blocks are aligned with the file system. If you don’t align them, this will kill your performance. If you use Puppet-Gluster, it’s all done for you. It also installs the GlusterFS packages, manages the peering and creates the volumes. I even have algorithms that figure out the optimal brick layout. You can also do chained layouts. If there are special brick layouts you may want to use, or try a weird layout, we can support that too.
People might think that GlusterFS is not good because it isn’t configured properly. But if you use Puppet-Gluster to setup GlusterFS, it helps to ensure you get a great out-of-the-box experience.