On 27 April 2014 02:57, veegee wrote: > They all fail. The only question is when. > > In my (long ago) prior corporate lifetime I used to tell users that all hard drives fail - somewhere between 10 years and 10 minutes after you first use them, and that you can't be sure which it will be. Quick look suggests that the oldest of the 15 USB connected external drives on this system is a 5 year old Seagate 1 TB. So far, as far as I know, it has behaved well. Time for some overall investigation I think. > I treat hard drives as disposable things that can spontaneously combust a= t > any time. He's reading my mind :-). I used occasionally to use an analogy for fun that gave a good enough idea of what to expect. This has been embellished somewhat as I go along :-). "Imagine a castle with a large banqueting hall. On one side there is an immense open fire that is stoked night and day. Occasionally the fire sends out showers of sparks of greater or lesser size. Occasionally sparks drif a long way. Sometimes a long long long way. Who can say how far they will drift today. On the far side of the large hall is a great sheet of super-fine vellum stretched paper stretched floor to ceiling on a frame. Scribes use ladders to ascend the wall and record the annals of the kingdom on the vellum. Every so often sparks from the fire reach the vellum sheet. Sometimes they produce no result. Sometimes they may a mark so minor as to be wholly irrelevant when reading back the related annals - the eye corrects for the mark unawares. Occasionally a small flareup and burn occurs and a word or sentence or a few scattered words may be lost. These may be able to be corrected. If a burn is too large the text may be rewritten elsewhere. Sometimes a large hole may occur. Very occasionally the whole sheet she goeth up in a flaming conflagration, senor. So, on another wall there is a duplicate sheet that ... Vellum sheets have been known to last for over a decade. Years is usual. The late Cedric the sorrowful copied out a whole new sheet after a conflagration and it was gone by lunchtime. Them's the breaks. I've seen a new HP hard drive last half a day. Sure, it must have had issues that should have been detectable. But data could still be lost on it= .. I've heard a new disk last under a week. Very loud while it lasted. For anything important, I use 3 drives in RAID 5 mode. I set up a > Linux server and use ext4 over mdadm RAID 5. > I don't use RAID - and it may well be a better idea that what I do use. I cross drive backup main day to day directories and keep 2 or 3 or sometimes more copies of associated groups of photo files, which are the main space consumer. Single "events" may have a master DVD but I'm tending to use USB memory sticks more and DVD capacity is too limiting. I've never got to using BlueRay. > > Seagate drives are cheap and quite fast for a large capacity. The other > brands may be more reliable on average, but they cost significantly more. Seagate and WD are comparable in price if bought when on special from lowest price sellers - which is the way I buy mine as price difference from walk-in-and-buy can be vast. My informal target is $NZ50/Terrabyte - best prices lately are somewhat up on that. 3TB are usually best MB/$ at present and 4TB will soon be similar. My experience and the report I cited suggest that Seagate are typically MUCH less reliable than WD, but also model dependant. I > > don't need my drives to last 10 years because they'll be outdated by then= .. > If a drive fails, I replace it with a larger, newer one and rebuild the > RAID array. I can use the extra space for whatever I want; LVM and mdadm > make it very easy to manage volumes. > My smallest external drive is 1 TB and largest is 3TB. 4TB soon. A failed 1TB is about 3.5% of total capacity and the duplicate of what it held that needed recopying would be copied from backup to a new 3TB. I dislike using systems that span volumes in ways that I do not have control and awareness of. That's just me. Others are happier to trust the machine to do a good job and allow it to make decisions which are somewhat opaque at the lower levels. Russell --=20 http://www.piclist.com/techref/piclist PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .