Contributed by jl on from the zip-to-my-lou dept.
It had been a few months since I had experimented with softraid in OpenBSD, so I though I would give it a try and see how far it had progressed. I have been using bioctl with hardware raid and was quickly won over by the elegance of the tools, and I looked forward to managing software raid with the same simplicity.
Just to be difficult, I dragged an old dual 450mhz Apple Power Mac out of storage and connected a USB and IDE iomega Zip drive to function as my raid volumes. Unfortunately SMP isn't supported by OpenBSD on the PPC platform yet, but everything else was detected and functioning correctly.
Softraid is enabled in current snapshots, so there was no need to compile a custom kernel.
The man page for softraid(4) is sufficient to get started, although I did add some suggested extra steps just to make sure.
The first step was to mark the zip drives as raid volumes. The disklabel(8) commands in the man page didn't work correctly for me on PPC because fdisk(8) creates an additional small MS-DOS volume, so I had to manually set the proper offset in disklabel for the RAID portion of the disk.
Once each zip disk had a RAID partition, I was able to use bioctl to create a new volume, and at that point it appeared as just another SCSI drive. I continued to follow the steps in the man page, mounted the RAID volume, and started writing and verifying data. Both RAID0 and RAID1 worked great, although I needed to use dd(1) to wipe out the beginning of zip disks to make sure all the softraid metadata was really gone before trying to re-use a disk in a new RAID set.
Of course, given my hardware choices, performance was terrible, but it hadn't corrupted data or caused a kernel panic. I was happy to see that the kernel now recognized the RAID metadata on the ZIP disks at boot and automatically assembled them for me. I even tried swapping the zip disks, and the kernel picked up the change and notified me that it had found a "roaming device". If I brought the system up with one disk ejected, it gracefully noticed the problem.
Unfortunately, softraid cannot yet recover from failure or automatically rebuild after an unclean shutdown, but I can see great signs of progress. I really look forward to a supported alternative to raidframe(4), and I'm excited to have the same management tools for both my hardware and software RAID. I'm also looking forward to seeing what unusual features the developers come up with, and the source tree is already leaving some interesting clues.
(Comments are closed)