On Thursday November 20, tomimo+linux-raid@ncircle.nullnet.fi wrote: > > > I had a mirrored pair that kept getting data corruption. It turned > > out that one of the drives had a bad bit in an internal buffer, and > > would occasionally return 1 for the 12th bit (or similar) of each > > sector even when it should be zero. Neither software raid or > > hardware raid would cope with that. > > How did you manage to figure out what was the reason for this > corruption ? I mean, does there exist a some sort of "generic" > way how to solve such errors, or was it just hard work in comparing > what got written to the different disks ... I think the generic solution is "Tear your hair out and wander around muttering to everyone how this couldn't possibly be happening and seeking sympathy". Like lots of problem solving, it is a case of gathering as much information as possible, staring at it for a while, and then explaining to someone who has absolutely no expertise in the field exactly why this collection of evidence is internally inconsistent and it just isn't possible. Usually that brings the answer out very quickly. (Some people say a brick wall will do, but I find a person works much better). One interesting aspect was that the drive with the problems was the one that it would preferentially rebuild the raid from. That tended to make transient errors more permanent as they would get written onto the second driver. I'm not sure if that helpped or hindered though. > > > Some people find that their IDE controller works fine until they try > > to use RAID1. RAID1 hits multiple discs concurrently a lot, and some > > (few, specific) ide controllers don't appear to cope. You probably > > would not get that with hardware raid. You can with software raid > > because it is a "whole-system" thing. > > Can you name the controllers which have had problems in this area ? > I'm having big problems in Linux 2.4 kernels IDE-stability with > heavy I/O (all disks are RAID1), perhaps either the Sil680 or > HPT374 are just problematic controllers ... ? (it would be just great > to solve all of these problems by buying new IDE-cards :) I'm afraid I haven't collected names. I avoid IDE raid like the plague, and just stick with SCSI - too many bad experiences. NOTE: This is not advice on my part to not use IDE. My experience is not sufficient to base a recommendation on. But for me, my budget allows SCSI, and it just makes me feel a lot safer. NeilBrown - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
