Bad Sector won't go away
 

On a Debian system - Unstable, AMD 10 6800K CPU, 4 GB RAM, 2 x 160GB SATA HDDS, 2 SATA DVD writers - I am getting the daily message:

When I run a short test
and then look at the result with
I always get the same result:
This is the partition table printed by fdisk:

I am interpreting this that my problem lies in sda5.

However, booting with Knoppix and running
only results in a message "Updating inode table" (or something similar) and the next day I get the SMART warning again.

Should e2fsck have cured this?

Help appreciated.

Your hard drive is a toast, it is not passing the test. Order a new one NOW. And make sure your backups are current.

Please post the output from "smartctl -A /dev/sda". The problem could be a simple as a single bad sector which just needs to be written to so that the drive can reallocate it to a spare sector. That is _only_ going to happen when a write to that sector occurs unless at some point the drive does manage to get a correct read from that sector and so can reallocate it on its own. Bad sectors that are pending reallocation will cause some offline tests to fail.

Assuming that the problem is just some small number of bad sectors, the Bad Block HOWTO shows the procedure for finding them, determining what file they are (or are not) part of, and making the drive reallocate them. If there are just a small number of bad sectors and this number is not increasing with time, then the drive is OK to use. There are various events such as vibration or power supply glitches that can cause a sector to become bad without being a warning of impending doom.

Good backups are, of course, always important. Drives can and do fail without warning.

Do not get confused, the test was not completed, it is a failure, the drive is dead.
Above is for sda, the info you should be looking at.

Feedback as requested:

I am assuming that Item 5 is the problem which is why I am reluctant to just dump the drive on this basis. sda5 is /home which is backed up daily.

@ Emerson

This is the output of the command you suggested. Does it look that bad that the drive needs to be dumped? OK, good excuse to get a bigger drive, meads I don't need to dump a lot of Beethoven to DVD :-)

The problem is #197, Current_Pending_Sector. That is just one bad sector, and the drive otherwise looks fine. A bad sector that is pending reallocation is visible to the OS (will cause an I/O error if read) and will cause the offline test to fail at that location. Follow the steps in the Bad Block HOWTO to get that sector reallocated. Parameter #5, Reallocated_Sector_Ct, should then increase to 2, and the offline tests should then pass. That drive hasn't been used much, just under 170 power-on hours, and you should expect it to have a normal lifetime.

The steps in the HOWTO aren't as hard as they look (it covers several different cases -- you will be concerned with just one), but if you don't want to do that, the ham-fisted approach would be to back up the files on the affected partition, clear the partition with "dd if=/dev/zero of=/dev/sda5 bs=64k", then remake the filesystem and restore the backup.

Of course if you just want a bigger disk, by all means go ahead and get one.

BTW, when you post output please use [CODE]...[/CODE] tags and not [QUOTE]...[/QUOTE] tags so that formatting is preserved.

The attributes look fine, except of course for the bad sector, which is not good. You could try zeroing the HDD like rknichols suggests as this may repair soft errors. Obviously backup before doing this.

Indeed, this will make your posts much more readable. I have fixed that for now.

Apologies anf thanks for the format fix.

I have tried following the Badblocks Howto but have run into a snag. Here follows a blow by blow account in the hope that somebody will point out where I went off the straight and narrow path.

Step 1: Find error:

Definitely at 144701458!

-----------------------------------------------------------------------------------------------
Step 2: Locate Partition where the error is:


Block number = 144701458 x 512 / 4096 = 18087682.25

18087682 must be in sda1


Step 3: Find Mount Point and fs type

looking in /etc/fstab I find:

No surprises here (I think)

Step 4: Confirm the Block Size:

Ok, 4096 as assumed earlier,

Step 5: Now to locate the inode:

I don't know what the error message means. I would appreciate being told what I am doing wrong!

LBA is in 512-byte sectors. "fdisk -u" gives addresses in 512-byte sectors. (The "Blocks" column shows 1024-byte blocks.) So, your bad block is in sda5, as you first suspected.

(144701458-39070143)/8 = 13203914.375

Block 13203914 of the filesystem, 3rd sector of that 4K block.

Thanks for this. However, moving along, I get:
Which rather puts a damper on the proceedings. icheck takes about half a minute to run, could it be timing out? I don't see how it can not find a block it has previously found with the testb command.

Could I trouble you a little longer?

Thanks.

That means that the block is used by filesystem metadata, probably by some currently free inodes. The only way I know of for finding which one is to runand page down through the listing until you see block numbers in that range, e.g.Unfortunately, the program will probably die from an I/O error at that point, but hopefully you will be able to see the "Inode table at ..." line and can confirm that the bad sector is within that inode table. The inodes in that sector pretty much have to be free or else your e2fsck would have died with an I/O error, so it should be safe to zero them. First, to be absolutely certain you have the right sector runIf you do get the expected I/O error from that, zero it by runningThat should make "smartctl -A /dev/sda" report "0" for the Current_Pending_Sector count, and the Reallocated_Sector_Ct will probably increase to "2". It would be best to run "e2fsck -f /dev/sda5" just to be sure you haven't stepped on something in use.

You did say you had backups for this filesystem, right? ;)

BackupPC, daily at 1500 :D