Impending HD failure?

[Robert S]

I'm starting to get these:

[Thistled]

I have been seeing these errors on 2 of my disks since installing gentoo back in 2008.
I thought it may have something to do with dual booting with windoze, as on one occasion I restarted my PC from Windoze str8 to Gentoo and there was a lock on the ntfs disk I was trying to mount. The solution was to shutdown windoze, then boot into Gentoo, and I would subsequently get access to the disk.
I tried defragmenting windoze to see if that would resolve it, but no joy.
The sectors always seem to be of the same size, and no increase over the years.
Just as long as you have made a backup of your important stuff, then I would not worry too much about this.
It sure as hell scared the ***t out of me when I first saw this info, but it has not escalated since the first warning, so I am not too worried.
_________________
Whatever you do, do it properly!

[Jaglover]

You should run something like this

[srs5694]

I strongly advise both of you to run a full SMART diagnostic on the disk. This can be done with tools like smartctl (text-mode), GSmartControl (GUI), or Palimpsest Disk Utility (SMART options are buried in a menu somewhere). IIRC, smartctl and Palimpsest are available in portage, but for some reason GSmartControl isn't. You might also be able to run a SMART test using a utility provided by the disk manufacturer, but that's likely to be written for Windows. This might be OK if you dual-boot, but on a Linux-only system, this could be problematic.

Unfortunately, SMART diagnostic results can be difficult to interpret. Some manufacturers put weird values in some fields that make things look worse than they are. Some fields are strangely named, and utilities often provide poor descriptions of what they mean. As a general rule, the GUI tools make the results easier to interpret than do the text-mode tools.

If the SMART tool gives you anything but "passed" for its overall assessment, you should probably replace the disk ASAP. Likewise if individual tests look troubling and you get confirmation from an expert that this reflects a real problem. The whole point of SMART is to detect disks that are just starting to flake out, so that you can replace the hardware before it fails entirely. It's possible to go for days, weeks, or even months with a disk that SMART says is problematic, but such disks are much more likely to go south very quickly than is a disk that gets a clean bill of health from a SMART test.

Edit: I posted just seconds after Jaglover. By "both of you" in my first paragraph, I'm referring to the first two posters.

[BillWho]

Robert S,

I've had similar errors on a disk for close to three years now. I have gentoo installed as test and break system so there's nothing important on it.

I saved the output of /usr/sbin/smartctl --log=error /dev/sdb and it still reports the exact same info today.

That disk could live another several years with no problems or it could crash and burn tomorrow.

If you have any critical data on it then for sure back it up - don't take any chances.

Good luck

[Thistled]

In my case all the disks which are reporting errors

[Mad Merlin]

Those errors are exactly what they sound like, a sector is unreadable on the hard drive. That sector might be part of your swap file (probably won't matter) or it could be part of your /boot/grub/grub.conf (not so good). Reads to that sector will fail. The next write made to that sector will cause the hard drive to transparently remap that sector to another spare sector and everything will be normal again.

Now, hard drives have a relatively small number of spare sectors (think dozens), and eventually it will run out. What happens after that is left as an exercise to the reader. Ideally, you will replace the drive before you are able to find out.

This might sound bad, but bad sectors are a fact of life, just as are dead pixels on your monitor, hard drives will deal with them just fine in small quantities. In general, if you see a small number of offline uncorrectable sectors and that number is not rising over time, the drive is probably fine. If you see a number that's steadily (or quickly) rising over time, toss the drive, it's going to eat your data.

Of course, I would point out that I've seen plenty of drives die completely out of the blue (SMART had no complaints right up until the drive's block device disappeared). Consequently, it's always a good time to test your backups.
_________________
Game! - Where the stick is mightier than the sword!

[Robert S]

Here's the output.

[Thistled]

[BillWho]

Thistled,

I don't believe that you can attribute the errors to winblows. I have a winblows installation on my disk and no errors are reported with smartctl.

[srs5694]

I agree with Mad Merlin: Back up your data and either replace the drive ASAP or be prepared to lose it suddenly.

One more point: SMART tools work with the disk hardware itself to detect problems. As such, SMART works at a much lower level than filesystem drivers. SMART can detect errors in parts of the disk that are unused -- unused parts of a filesystem or even gaps between partitions. Thus, you can spend all day running fsck in Linux or defragmenting files in Windows and there's no guarantee that you'll touch the affected sectors. Likewise if the bad sectors are in the middle of a big file that happens not to be adjusted by a defragment operation.

The best way to ensure that you do something with a sector that's going bad is to do a raw write operation to the whole disk, as in:

[Thistled]

I am a little confused by all of this. This 1st disk is my Winblows disk, and is barely used by Linux, but I can mount it if I want to install any apps using Wine.

[Jaglover]

Alright, have you run self-test on this drive? Did it finish? If the drive is bad the test usually will not accomplish.
Below is a sample of a healthy drive.

[Thistled]

Well palimpsest reports I have 12 bad sectors on both sdb and sdc

and

[Hu]

[Thistled]

But this has been like this since the day I bought the disk.
Palimpsest has always reported the current pending sector error.
Like I said in earlier posts, all my important stuff is backed up on my server, I am kind of taking that same approach as BillWho.