David T-G said the following on 10/24/2013 07:03 AM:
...and then Anton Aylward said... % % David T-G said the following on 10/23/2013 11:24 AM: % > % >Oh, and one other question... Do I do my RAID via LVM, or other? % % Learn more before you ask that question, as right now its clear a reply % yes/no would make no sense.
Yeah, I know :-(
% First, LVM can stripe and can do mirroring WITHOUT RAID.
Wait a second... What are striping and mirroring except implementation of the Redundant Array of Inexpensive Disks that is the RAID concept?
That's exactly what I mean when I say this isn't a yes/no issue. You're trying to fit things into mental bins without having enough bins/categories and without adequate 'filters' on each bin to say 'no, that doesn't belong here'. RAID is a collection of techniques that may or may not make use of a number of drives in any one of a number of various ways. It helps to know a bit of coding theory and understand Hamming, and also understand what error rates really are and mean, and how different handlers deal with errors. For example, even in ancient history each sector on a DASD had a CRC checksum. CRC - thank you Hamming - allowed the detection AND CORRECTION or errors in the read of the sector, Or in the read of the CRC. To a point. But Back Then, the controllers didn't make use of it, so I wrote a handler for PDP-11 UNIX V6/V7 that did for NorTel. If the errors persisted - that is they weren't statistical fluctuations - I wrote the corrected/best-effort sector to a reserved area and set up a look-aside table entry. Fast Forward and we now have the concept embedded in the controller strapped on every drive you buy. So the effective error rate is way lower than the error rate or the platter media. Engineering is like that - engineers find ways round physical limits. Just watch us defeat the light-speed barrier :-) So RAID-5 made sense of using Hamming checks to correct even those errors. Now go read the articles on why RAID-5 makes no sense with large capacity drives. That's RAID-5 Mirroring? We had mirroring long before the term RAID came into use, back when DASD was still expensive. Heck, I know businesses that mirror their whole site, not just a disk! "Hot Backup". Striping isn't new either. IIR correctly it was even done before DASD was available, done with tape drives. I've never seen that but I've read of it. Yes, once people used computers without disk drives! And yes, they pioneered most of the techniques we use today. With primitive hardware, slow CPUs and a lot less memory. We've quite a few Greybeards on this list. And some of even read about History.
% % Second, putting LVM on top of, say RAID 5 implemented 'normally' can % make it more manageable. We've mentioned that in the past - check the % archives.
Yeah, but the most I'll have is mirrored; I don't have the devices for checksumming. I hate that it's only 50% efficiency, but it's what's available.
More to the point its simple in concept and simple to implement and simple to do recovery. You don't need RAID software - LVM can do it, and if you don't want it done at the OS level there are hardware cards that can do it for you that are pretty cheap. One large AIX multiprocessor system I worked on had, for each (multicore) processor, a (comparative) small mirrored disk pair for just the OS. The data was in the next room, wall-to-wall racks of RAID.
% % But if I were you, I'd take a disk, use parted or whatever to create a ... % but is more than enough to do an install.
That sounds lovely, but it's completely orthogonal to what I want. All of the spinning space is for data and has nothing to do with the OS, which is on a uSD USB drive.
So? You spinning stuff lives under /data and its on drives that implement techniques for speed and reliability - which can be done with LVM. LVM is about resource management. Add another drive - any size, when budget permits or when you find one in the Closet of Anxieties, add it to the volume group, and LVM will make use of it. One drive turns flaky? Tell LVM to remove it from the volume group then pull it. -- The deepest sin against the human mind is to believe things without evidence. Thomas H. Huxley -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org