Fskc problem

[Radim Jarosek]

Hi,

I have a problem after restart server (power outage) with fsck. I got error message that the fsck failed and I started it manually.
I'm not sure what else I can try it. Could you give me a hint ?


Thanks in advance

Radim

**********************************************

System : hp-ux 10.20
Filesystem : hfs

**********************************************

fsck -F hfs -y /dev/rdsk/c3t5d0
** /dev/rdsk/c3t5d0
** Last Mounted on /backup
** Phase 1 - Check Blocks and Sizes

CANNOT READ: BLK 2555936
CONTINUE? yes


CANNOT READ: BLK 2555936
CONTINUE? yes


CANNOT READ: BLK 2555936
CONTINUE? yes

FAILED READ OF BLOCK #2555936, RETRIED 2 TIMES
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
SUMMARY INFORMATION (INODE FREE) BAD
BAD CYLINDER GROUPS
FIX? yes

** Phase 6 - Salvage Cylinder Groups
83868 files, 0 icont, 6761194 used, 10637090 free (19482 frags, 1327201 blocks)
DISK MEDIA PROBLEMS ENCOUNTERED!
BAD BLOCKS WERE FOUND ON THE DISK.
***** FILE SYSTEM IS NOT CLEAN -- DISK MEDIA PROBLEMS ENCOUNTERED *****

***** FILE SYSTEM WAS MODIFIED *****

[Robert-Jan Goossens]

Hi Radim,

Hope you have a good backup. I think the only option is to recreate your filesystem and recover the data.

Hope this helps.
Robert-Jan

[Julio Yamawaki]

Hi,

probably you have some physical problems and need to use mediainit.

[Shaikh Imran]

Hi,
Your disk has developed physical bad blocks.
Better take the backup and get it replaced.

Regards

[Radim Jarosek]

Hi,

I managed to mount file system (with force option) and now I'm copying data to another disk and then I will try to recreate file system.

Radim

[Roberto Polli]

Your pbm seems due to a broken HD, anyway
I sometimes experience problems of that kind : it happens on a LVM stripe made up of 2 internal hd and 2 external hd (well, it's not an idea of mine!!!).

The problem was related to scsi cable that once a year should be changed.

To restore your filesystem without rebuilding it from scratch you could try to replace the bad superblock with a good one (find into /etc/sbtab a good superblock) and do a
fsck -b GOOD_SB DEVICE


Pax, R.

[Radim Jarosek]

Hi Roberto,

how do I find a good superblock ?


Radim

[Bill Hassell]

The first available superblock is always at 16. The rest are created at block numbers determined by the size of the lvol. However, if the first alternate block is also bad, the disk is probably unusable. fsck cannot fix bad spots on the disk, only the structure and login of the filesystem itself. A force-mount can be done but be prepared for the possibility of a system crash in case the structure of the directory badly corrupted.

[KapilRaj]

/var/adm/sbtab will have a list

Kaps

[doug hosking]

When errors like this happen right after a power failure and appear to be limited to a single block I would think about the possibility that some block(s) got written with bad CRC/checksums or equivalent. I've occasionally seen this happen after dirty power failures, despite the protections that are supposed to be designed into modern disk drives.

What follows is a 'use at your own risk' method that sometimes saves the disk. I don't recommend it in cases where you have reason to believe that there are physically bad blocks on the disk (as opposed to just blocks with invalid checksums because power died in the middle of a write).

Sometimes you can recover if you dd the raw disk to itself from single user mode, including the options conv=sync and conv=noerror on the dd command line.

This should force the block with bad CRC to be rewritten as a block of zeroes. While this is still not a valid disk format, fsck is then able to complete the read and may be able to correct the problem.

Again, use at your own risk, preferably after you've saved whatever data you can by other means. It's worked for me a few times when I lived in thunderstorm country without a UPS to stabilize the power.