Openfiler is supposed to be an enterprise network storage operating system, and it is very capable, but in many ways it’s an exercise in frustration.
Rarely do our little IT empires grow in an orderly, cruft-free fashion, but rather they acquire new bits and pieces like old shipwrecks acquire barnacles. Barnacles acquire more barnacles, and mussels, limpets, anemones, oysters, and gosh knows what-all, and it all bonds together in a tough aggregate that’s nearly impossible to separate without damaging somebody.
Our datacenters grow in the same manner, and even with the best intentions and care rip-and-replace is rarely a feasible option, because just like our friends the barnacles when we change one thing it affects other things. Anyway we’re at the point where we don’t need fabulous new revolutionary technologies because Linux already has everything we need: networking, storage, services, virtualization, cloudy stuff… I think our most pressing need is for good management tools.
And so Openfiler 2.99 aims to be your one-stop storage management shop, a sleek lean network storage operating system with a nice Web control panel. Openfiler supports all manner of networking protocols:
- NFSv3 and 4
- CIFS/SMB
- FTP
- WebDAV
- HTTP.
And storage protocols:
- iSCSI
- Generic Linux SCSI
- Fibre Channel
- InfiniBand.
It offers snapshots and scheduled snapshots, live volume migration and replication, and fine-grained access controls on shares and quotas. It includes Cluster 3.0, DRBD, and Heartbeat for high-availability. Some other nice features are UPS (uninterruptible power supply) management, monitoring and notifications, and a Java-powered SSH console. (I know, just use the real console, but some users will like it.) The Web administration GUI is comprehensive and fairly well-organized.
The Ext 2, Ext 3, and XFS filesystems are supported, but oddly not Ext4, which seems like a natural for a storage operating system, since it supports very large storage volumes.
Some advanced plugins that are available only by subscription are the Advanced iSCSI Target Plugin ($1,295.80), and the Fibre Channel Target Plugin $1,295.80). Both are proprietary, and both are one-time, perpetual, per-instance licenses. You get some installation and migration help, and then additional support is extra.
Installation Gotcha
There are installation walkthroughs for both the text and the graphical installers. Be sure to review these because they contain the default login, and manual partitioning instructions. Do not let the installer automatically partition for you, because you won’t be able to create volume groups from the Web GUI.
Another gotcha is the screenshots don’t match what you’ll see in the installer, because they are for an older release.
Web GUI
The Web GUI is pretty nice and, in theory anyway, makes it fairly easy to set up some of the more difficult configurations such as RAID arrays, high availability, LDAP user authentication, or Windows Active Directory authentication. And it does live up to its promise of easier administration, and handles a lot of these tasks competently. But not all. For example, I thought it would be fun to pointy-click my way to an HA setup for once, instead of hassling with text configuration files. Usually I prefer text configurations because they are faster than wading through a GUI maze, and they’re easy to replicate, but Heartbeat’s ha.cf is a complex beast. In Openfiler it appears to be as easy as first visiting the Services tab and starting up Cluster, and then going to the Cluster Manager tab.
But no. Cluster does not start from the Services page. It just sits there and looks at me. So, after a fair bit of Web-searching I learned that is by design, because “Advanced features such as high availability (HA) and block replication, including WAN replication capability for remote disaster recovery (DR), support require an enterprise cluster or SME cluster subscription.” Now, I’m no ace Web programmer, but I do know how to put a sentence on a Web page that says “To activate this feature, please do such and so.” Which surely is preferable to leaving potential customers in the dark and thinking either they are missing something, or the product is defective.
No Documentation
Which brings me to my main gripe with Openfiler, and that is the lack of useful documentation, and what little they have is disorganized. There are no READMEs, no changelogs, no news, no blogs, no release announcements. When you click the “Release notes” button during installation all you get is a message that says “Release notes are missing.” The download page claims to offer multiple download images: an ISO for a hard drive installation, and several pre-fab virtual machine images. But there is only an ISO for the current release, 2.99, which was released in April 2011. There are multiple VMs for the 2.3 release, which is from February 2009.
Openfiler offers both community and commercial support. Their IRC channel and user forums are featured prominently on their front page, but they aren’t all that helpful. I had the IRC channel open for several days, and even though several dozen users were logged in it was very quiet. The forums have more questions than answers. So what is the puzzled admin to do? You can purchase an administration manual for $51.83. What do you get for your fifty-some dollars? I have no idea because there is no table of contents or overview, it doesn’t say how many pages it has, or even which version it covers.
Openfiler is based on rPath Linux. rPath is designed for easy customization and custom builds, and especially for building software appliances. The main difference for the hardworking system administrator is it uses the Conary package system, which is completely different from all the other Linux package managers, so you’ll have a whole new package manager to learn. The Web GUI does not include a package manager, though it does have a one-click system updater.
I’m all for software developers getting paid, and Openfiler’s pricing is more than fair. Support options range from about $1,000 to $5,500 per year, which are not out of line for a high-end enterprise product. Me, I wish there were fewer mysteries for getting product and release information, and for testing basic functionality. I’m more likely to become a paying customer when my first experiences with the product are pleasant and productive.
