Red Hat Chief Architect (and top presenter) Thomas Cameron’s SELinux for mere mortals breakout sessions are legendary. This year, Thomas decided to do for Gluster what he’s done for SELinux. Thus, the Red Hat Storage for Mere Mortals talk was born. This session offered a high-level overview of an install (from scratch) of Red Hat Gluster Storage 3 on Red Hat Enterprise Linux 7. He began with a quick history and architecture of Gluster, the upstream community project.
Thomas then built 4 demo virtual machines (VMs) from kickstart with @base, the minimum functional setting to get started. Though Red Hat Gluster Storage can be administered from a user interface (UI), Cameron focused on demonstrating functionality from the command-line tools, with NFS as the the mounting protocol (Gluster Storage also supports CIFS, Swift, and native GlusterFS).
Thomas noted that, on Red Hat Gluster Storage 3, SELinux needed to be disabled. He recommended that it be set to permissive instead so that you can watch everything in the logs. Disabling will not be necessary with version 3.1, as all of the components are now labeled in the targeted policy.
Once mount point /export/gluster (an arbitrary name, you can call your mounts whatever you like) was created on all nodes, Cameron ran ‘gluster peer probe’ for each node. Once a node has been probed, you do not have to run the command identifying that node again–the other peers will be aware of the new server. He then reviewed the 3 current and supported volumes types:
- Distributed Volumes
This is the most rudimentary configuration. All files are evenly distributed across the trusted storage pool. The benefit of a distribute volume is that it is fast. The downside is that there is no redundancy, if a peer dies then all files on that particular server are lost.
- Replicated Volumes
All files are written to each peer in the trusted storage pool. The benefit is redundancy, however, there is a performance hit.
- Distributed Replicated Volumes
This volume improves read performance while providing redundancy. Cameron warns that this is expensive, as it requires more servers than the previous 2 configurations.
If your laptop is capable of running 4 virtual machines, then you can test these volume types at home.