forum.excito.org

Hi,

I recently (well, not recently, but I recently discovered it) ran into some bad sectors in my bubba. I don't know why this happened or how, and I have thought that drives are supposed to relocate bad sectors behind the scenes. Well, either way, I now have some holes in the drive (and filesystem) that seem to be writable but not readable. Also, trying to read them causes SATA link resets and it in turn makes B3 extremely unresponsive...

During an attempt to fix the issue (e2fsck -c) I have sadly managed to make my b3 unbootable, presumably due to unfixable filesystem errors in /home (e2fsck failed to relocate a bitmap that now collides with blocks allocated into the badblock inode). Well, sad, that. The next step was the rescue image. First attempt failed, so I thought I must have made a mistake in creating the USB disk, started over with mkfs.vfat, etc. and on the second try, forgot to edit bubba.cfg and had my /home formatted. Well, worse thing happened, but I would like to point out that the default of "erase everything" is a tad bit dangerous. (Yes, I have mostly everything mirrored somewhere, but it's still quite annoying.)

Anyway, back to the problem. Of course, the reinstall happily wiped the bad block information from the filesystem, so I am at square one (with the benefit of not having to deal with corrupted files that are all gone anyway). How should I go about this? Right now I am running badblocks, stashing the list in a file, so at some point I can try to do a e2fsck -l and hope it goes better than the last time.

Do you think I should try to get the drive replaced? (It is the original 1TB drive that came with B3.) Or should I just keep around the list of bad blocks and hope it doesn't get worse? Is there a better filesystem (say, ext4?) that would cope better with bad blocks than the default ext3? (I most importantly mean that even if the filesystem was mostly usable for normal operation, e2fsck persistently died with bitmap relocation error).

Well, I have mostly restored things to their original state, after marking the bad blocks with fsck. So far, things seem to hold together.

And more bad sectors appeared. That's so very inconvenient. Anyone knows what's the warranty policy? I don't like the idea of being without a B3 for a month or more... (Nor I like the idea of losing a few hundred sectors a week...)

you really should contact support@excito.com for this matter

Ok, I am giving it a last chance, running a read-write badblocks test (which will take a couple days, it seems). I really hate warranty processes. So if a read-write scan gives a final list and no new sectors go bad I'm happy enough. I'm wondering if a sudden power loss could have caused this? Presumably, PCs have big power sources that give power long enough for the disk to park properly? (I have always assumed that modern disks handle power loss gracefully, but never thought whether it was a feature of the disk itself or it required cooperation from the power source.)

well, having same issue with my lvm partition (bubba home) - or file system table corrupt - not sure yet. but the read write test you mention would take in my case - on a 1tb hdd and a lvm buba home partition of 989 gb (1 gb swap, 10 gb bubba_root) knowing 1% = 1h maens 100% = 100h - more than 4 days.

this issue began on my behalf firstly with a power loss. i then had the chance that bubba came up after 1h 30min blinking blue led on 4hz. and now the second time was due to unplugging the power cable directly because pushing the button on the back pannel didn't shut down the system.

I have read up a bit on how drives handle power failures, and this should, on reasonably modern drives, never cause trouble. The mechanism supposedly uses the rotational inertia of the platter to feed power into enough of the system to park the heads safely. We had another power outage today though, and I have noticed SATA reset messages in dmesg again, even though yesterday I had a completely clean filesystem with a freshly compiled list of bad blocks from the read-write test. If this is indeed related to power losses, I'd say something is wrong with the drive. This shouldn't be happening.

@mornfall, sorry that you are having so much problems. You are right, bad blocks almost never happen on our systems (ext3 is good enough to handle these things normally) so you do have a bad drive. Just email support@excito.com and we'll replace your drive. Refer to this thread in your support ticket so you won't have to explain it all twice.