On 04/17/17 19:56, Ian Watts wrote: > Hello, > > I'm planning on replacing an old fileserver that has a single 1T drive > with something a little newer having 3T of space. I have two 3T drives > and have installed OpenBSD 6.0 to both as a softraid mirror. Works well > and I simulated a drive failure by shutting it down, removing a drive, > and rebooting. The drive has been re-installed and it is now rebuilding > the mirror. After 17 hours it is 24% complete, so it'll be about three > days to complete. The system is: > > AMD E2-3200 2.40 GHz > 4G RAM > 2 x 3T Seagate Barracuda 7200rpm SATA > > With this much disk space, should I be looking at another way of > achieving data redundancy? The goal is to increase redundancy of the > data and the mirror would be periodically backed up to another server in > a different building. My only concern here is the suitability of the > softraid mirror for a large filesystem. I've thought of using the > second drive as a backup and rsync'ing it nightly, but then failure of > the primary drive would mean more downtime before it's operational > again. A long rebuild time isn't a major problem; just want to make > sure I'm not overlooking a more sensible option. > > FWIW, I used the following info to get set up: > > https://www.openbsd.org/faq/faq14.html#softraidDI > http://openbsd-archive.7691.n7.nabble.com/Large-3TB-HDD-support-td95308.html > > Thanks, > > -- Ian
Keep in mind, it's easy to say and now trivial to buy "3TB disks", and therefore, it's easy to forget that it is a SNOOTLOAD of data. Three days to mirror 3TB isn't out of line for some HW mirroring systems I've worked with, and much faster than many. Still...verify that you are running with an ahci(4) controller (sd(4) disks), not a pciide(4) controller (wd(4) disks) (though at one point, I don't think it was possible to have wd(4) disks that big, not sure if that's still true. And I suspect if you were running wd(4), it might be weeks, not days). And yes, when you have a three TB of data and a three day rebuild period, the possibility of a second disk failure during rebuild is definitely not zero, so yes, I'd suggest *considering* some alternative ways to achieve data security. * Three disk RAID1? (a REALLY good idea) * Checksumming "static" files? * rsync'ing data between stand-alone disks? (IF you can restrict the amount of data you have to have rsync look at, you can sync a LOT of data very quickly) * "Chunk" (or partition) your data as best you can, so you can mount blocks of storage Read Only, as "full and unchanging" (note lack of questionmark -- you want to do this if at all possible) (chunk your data, but NOT your RAID partitions -- last thing you want to get stuck doing is remirroring multiple RAID partitions on one disk at the same time!) * Something else relevant to your situation? Nick.
