Problem with fsck

[Stohler Andreas]

When we do a fsck on a lvol we got the following result:
fsck /dev/vg00/lvol8
log replay in progress
vxfs fsck: fsck read failure bno = 43180, off = 0, len = 1024
full file system check required, exiting ..

Then we do a
fsck -F vxfs -o full,nolog /dev/vg00/lvol8

Result:
pass0 - checking structural files
pass1 - checking inode sanity and blocks
vxfs fsck: fsck read failure bno = 43180, off = 0, len = 8192
file system check failure, aborting ...

Does anybody know why it's aborting ? Is my disk crashed or can I fix this, ? How?

[Michael Tully]

Hi,

I suspect you have a disk that could be failing. Suggest you have a look at /var/adm/syslog/syslog.log file for any messages. You could run some disk exercise tools to see if it reports anything.
You could try stm/cstm/mstm/xstm and exercise the disk, or even use 'dd'

e.g.
# dd if=/dev/rdsk/c1t2d0 of=/dev/null bs=1024k

Regards
Michael

[James Murtagh]

Hi Stohler,

Its aborting as it cannot garauntee the consistency of the filesystem due to some problem reading block 43180. The first thing to check would be the syslog and message buffer for an IO/VxFS errors:

# grep vmunix /var/adm/syslog/syslog.log
# dmesg

You can also try to physically read from the disk partition to check for any IO problems:

# dd if=/dev/vg00/rlvol8 of=/dev/null bs=51k

You can adjust the blocksize depending on the size of the logical volume. You may also check the physical block using fsdb:

# fsdb -F vxfs /dev/vg00/rlvol8
> 43180 b; p 100 x

This will read 100 bytes in hex from that location. Use d to quit back to the shell. You can also check for stale extents on the logical volume:

# lvdisplay -v /dev/vg00/lvol8|grep -i stale

Depending what commands report the errors you will probably have to replace the disk or recreate the filesystem. If you post some of the output I can help decide which one.

Other things to be aware of:

i. Ensure you are properly patched. You never mentioned the OS but if at 10.20 or 11.00 ensure you have the LITS patches installed (PHKL_16751/PHKL_18543)
ii. If by some chance the logical volume resides on a disk array you must check the recommended bad block reallocation policy for that array. All high end arrays, XP, EMC for example you must have bad block reallocation set to NONE. You can see this from lvdisplay.

Regards,

James.

[Stohler Andreas]

Hi Michael,

Unfortunately in my defect
lvol is /var. So I can't
check syslog.
Also tools like stm won't
start.

[Stohler Andreas]

I think my disk is
gone:

dd if=/dev/vg00/rlvol8 of=/dev/null bs=51k
dd read error: I/O error
846+0 records in
846+0 records out

Or any other ideas?

PS: I'm havin HP-UX11.0 installed

[James Murtagh]

Hi Stohler,

Sorry, can I take from the last comment you are just installing 11.00 onto this disk?

If this is not the case what does the lvdisplay output show? Is the disk mirrored?

In any event your disk appears to be faulty and should be replaced. If it is mirrored you may be able to reduce this mirror and fsck the filesystem on the other disk as a workaround until you get it replaced.

Regards,

James.