On 29/12/16 07:35, jdd wrote:
Le 29/12/2016 à 02:28, David C. Rankin a écrit :
On 12/24/2016 02:31 AM, jdd wrote:
right now this is a test config for which I use stock hardware, I may as well completely drop raid if I find it too complicated or too expensive, but I had never the occasion to play with it :-))
raid is never to complicated or expensive. mdadm is absolutely bullet-proof
there is an other thread around that shows it's not bullet proof and strong advisory on the raid wiki not to use consumer grade disk. This alone makes it expensive. I can use several disks because I already have then in stock, so they are not free but cheap.
The reason you shouldn't use consumer disks is that you can NOT alter the time-out on them, and the disk defaults interact badly with the linux defaults. That said, the difference in price between NAS and consumer disks isn't much - I think a 3TB Barracuda (bad choice) is currently £70, while a WD Red (good choice) is £100. The thing with a consumer grade disk, even if you're not doing raid, is that a problem with it will cause your computer to appear to hang. So desktop disks aren't really even suitable for a desktop! :=)
it's a bit complicated because the wiki says very hard than one can kill it's config trying to recover. also, the commands to regain access to the disk in case of big failure (for example mobo failure) is more complicated than the usual rescue through bind mount.
Fact is than in more than 10 years of server use, I had only *one* disk failure (last summer) and I have no real "instant recovery" needs, so recovering from a backup is possible.
Having seen several disk failures, I'd much rather try and recover a broken raid, than a broken disk. I think I've managed to recover a maximum of two broken disks, and I've seen a few more than that that I couldn't recover.
I'm trying raid mostly as a game and to use my sleeping hardware :-)
I've got a raid mirror, because backups would be a lot of hassle. I do try and make sure anything important is flushed to DVD, but my home server has a 2TB /home partition, and it's over half full ...
and I know openSUSE community (and open source community at large) is very friendly :-)
As is the linux raid community :-) That warning on the raid page is more really to scare people into coming to the list sooner rather than later. Unfortunately there's some really bad raid advice out there on the net that people fall foul of. And if somebody comes to the list BEFORE they try anything "dangerous" on their array, we can usually recover it. It's when they do something "stupid" because the net told them to, that it gets difficult. And the wiki is meant to be pretty comprehensive because, well, we see a lot of the tricky recoveries, and you'll notice that most of the pages say "gather this information and post it to the list", because then hopefully we can come straight back and say - "got it, do this, and your array should be back fine". The next cases that need writing up are where something has trashed the disk headers (GPT etc). There've been several successful array recoveries recently where that has happened ... Cheers, Wol -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org