August 8, 201213 yr First, go back up your important data and keep a reliable copy off site. No that said, most of us have some degree of trepidation regarding a 2-drive loss.... particularly nerve-biting is the unprotected period after a drive failure, of a dreaded red ball on a second drive during rebuilding of the first failed drive. Obviously a failure of the second drive to be readable at all means you are toast.... so I am not talking about a really "dead" drive. I'm talking about a hard read error where the rest of the drive is readable. (IMX, most (90%) failures in a live running system have been drives that were still readable and not "dead." When I've seen a "dead" drive happen it was much more common at power up rather than when the system was already up and running.) I can see several ways to react to that: 1) just log the locus of the sector(s) that can't be rebuilt and continue... and perhaps afterwards, report the file(s) that occupied the sectors that could not be rebuilt. 2) try to read the other drives, including the first failed drive, and rebuild that single sector on the second failed drive. This is the trickiest since you have to have the first failed drive still attached (or attachable) to the system. 3) suspend the rebuild, and let you power-cycle the machine, and perhaps fix a bad connector or power issue, and let the rebuild continue. 4) use some of the techniques from things like ddrescue to try to recover the bad sector(s) from either one of the failed drives after the fact... so skip errors and get the maximum rebuild as fast as possible, and then go back and try to "fill in" the un-rebuildable sectors. It seems not a Herculean task that when an error is encountered during a rebuild to at least check the drive that had the error during rebuild to see of it is "dead" and rebuilding can not continue, versus it had a reading error for one location, but much of the rest of the drive is still readable so rebuilding can continue with likely success. Thoughts?
August 8, 201213 yr DD rescue saved my butt when I hit a number of unreadable sectors on one drive. By doing a forward copy, then a reverse copy, I was able to recover all except 1 bad sector. It means taking the array off line and using ddrescue from one drive to another. It can take a few days to do this. At the very least it will fill in the missing sectors with 0x00 so you can recover with 'some' corruption. I think it's critical these days to do a md5sum/md5deep on each drive and store that file somewhere so you can detect corruptions and restore what may be lost. Saved my butt with a recent drive failure also.
August 9, 201213 yr Author Another good one is HDDRegenerator. http://www.dposoft.net/ I use ddrescue first, but HDDRegen has fixed drives that ddrescue could not get data from.
Archived
This topic is now archived and is closed to further replies.