Mike Marion wrote: > > Not that this will help too much, but IIRC RH defaults to running the slocate > update at 4am. Perhaps the huge increase in disk I/O is somehow triggering > the problem. > > Try running the slocate update by hand and see if that causes it to get kicked > out. That might help narrow down that heavy, sustained I/O is trigging the > problem in scd. > Yep, slocate and a bunch of other stuff in cron.daily goes off at 4 A.M. I ran slocate by hand and it did indeed reproduce the problem. I added the md0 drive mount point to the exclusion list in cron.daily/slocate.cron. That should patch over the problem, but doesn't really solve it. There's still the question why the RAID error recovery is so weak. Gérard Roudier wrote: > > I suggest you to check how the 'Read-write recovery page' is set up on > your disk. You must ensure that the ARRE bit, at least, is set. But this > will not magically make your disk reassign the faulty block. In fact, the > disk will reassign the block only if it has been able to recover from the > error and the behaviour is governed by the other infos in the page. In no > case, the drive will decide by itself to reassign the block and to copy > corrupted data to the new block. I think I follow what you are saying. The 'Read-write recovery page' is one of the status inquiry pages defined by the SCSI standard if I remember correctly. I'll look for a tool that lets me examine that kind of stuff. I think I saw something in the SCSI list not too long ago that might help. > If you have time to play, you can try, for example, to read the block > using 'dd', expect to get the error, and rewrite it immediately. This has > been reported to make the trick. > > If you haven't time to play with all that stuff (btw, I donnot have too), > just try to reformat the drive. You must just be aware that if the > drive is switched off during this operation, it may get unusable forever. Hmm. If I understand you correctly, you are suggesting that the sector with the error might be corrected if I use dd to read it. That would work best after the drive has been removed from the RAID set. I'll try that. However, the block is being read every night as part of the slocate job and being rewritten fairly often when the RAID set is reconstructed after a raidhotadd. To reformat the drive I'd need to find a linux low level formatting tool or move the drive to another machine which has a BIOS low level formatter. I'll pass on that for the moment. (I've done this a couple times with other machines and I feel much more comfortable with the machine connected to a good, well charged UPS when I do it.) [EMAIL PROTECTED] - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [EMAIL PROTECTED]
