FSCK Fails

[brian_31]

Team:

one of my lvols's is not mounting. the disks seem to be ok. I can see them in ioscan, but..when i do full fsck with a -y it fails. Looks like the filesystem is corrupted. Do i have any other option? I have a backup which is 2 weeks old but i have lost some critical data. Could someone help me here? Has someone used fsdb? i need this data. Please help.

Thanks

Brian.

[Michael Tully]

I suspect you could have a dead disk somewhere. Just because they show up in ioscan does not mean they are ok.
Two things first:

Do a vgdisplay -v /dev/vgxxx to see if all disks get reported. Look in the CUR PV/ACT PV rows.
e.g.

VG Name /dev/omni
VG Write Access read/write
VG Status available
Max LV 255
Cur LV 1
Open LV 1
Max PV 128
Cur PV 2
Act PV 2
Max PE per PV 2157
VGDA 4
PE Size (Mbytes) 4
Total PE 4314
Alloc PE 4314
Free PE 0
Total PVG 0
Total Spare PVs 0
Total Spare PVs in use 0

You could also do a diskinfo /dev/rdsk/cXtYDz on the disk

[brian_31]

I know the disks are good for sure. vgdisplay is fine and also the diskinfo. The volume group that this LUN belongs (we are using VA7100)is fine. It is only the filesystem that is corrupt.

Thanks

Brian

[brian_31]

this is the exact error
fsck -F vxfs -o full -y /dev/vg02/rlvfs
vxfs fsck: fsck read failure bno = 2834552, off = 0, len = 8192

Any ideas pl.help

Thanks

Brian.

[Ermin Borovac]

Next thing I would try is check disks in this logical volume with dd.

# lvdisplay -v /dev/vg02/lvfs | more

and for each disk

# dd if=/dev/dsk/c#t#d# of=/dev/null bs=256k

If it returns I/O error you most likely have disk that need to be replaced.

[Bill Hassell]

What is the error from fsck? If is says something about I/O error or unable to read block ###, then you indeed have a damaged disk. If you have the online diagnostics installed, syslog will be full of error messages. Look at the dmesg command. ioscan and diskinfo do not provide any integrity information, just a confirmation that the disk can ID itself. Use the dd test to read the disks. If there is an I/O error, fsck cannot fix it. This is the reason to always have DiskMirroring installed.

[brian_31]

This is the exact error
fsck -F vxfs -o full -y /dev/vg02/rlvfs
vxfs fsck: fsck read failure bno = 2834552, off = 0, len = 8192

Also tried the dd if=/dev/dsk/c12t0d2 of=/dev/null bs=256k . It returned 8192+0 records in 8192+0 records out

Thanks

Brian.

[Bill Hassell]

Since there is no I/O error message, this looks like an lseek error. Did you recently try to reduce the size of this filesystem? lvreduce is *always* 100% destructive because the directory elements are all over the disk. if you did run lvreduce, you can probably recover the filesystem by using lvextend to put the space back. There is an extra-cost product that allows filesystems to be resized called Online JFS.

[brian_31]

Bill:

I have online JFS. How shd i lvextend? the filesystem fails to mount? so i shd just lvextend?

Thanks

Brian

[Bill Hassell]

You need to know the original size of the logical volume in order to use lvextend. lvextend and lvreduce change the size of the lvol without regard to any data on the volume. Using lvextend assumes that the original volume was contiguous and that other lvol changes in the volume group have not captured the space. This will be a trial and error procedure if you don't know what happened to the lvol or you don't have a bdf or vgdisplay for the system prior to the problem.