Determine root cause of corrupted files on one file system

[Geoff Wild]

Have begun migrating to an EMC DMX from and 8830 EMC Sym.
Everything went fineno issues....however, sometime between 16:00 and 08:30 the next
morning, files in the /sapmnt/XXX filesystem became corrupt:
Here's the cksum from the "new" filesystem:
# cksum /sapmnt/XXX/exe/R3trans
2197931645 5630688 /sapmnt/XXX/exe/R3trans
Here's the cksum from the old:
# cksum /sapmnt2/exe/R3trans
2580146740 5630688 /sapmnt2/exe/R3trans

Nothing in syslog, no errors on DMX frame...

Need help determining the root cause...

Of course, R3trans will not run - bu if I re-copy it - it is fine...this is not limited to just executable/binary files - have also found corrupt text files - it has only affected this 1 lvol - all others (including db's) are fine....

Thanks...Geoff

[George A Bodnar]

I would suggest double-checking the device of the new volume to make sure you don't have a situation like multiple hosts accessing this volume or are these 1-to-1 type mappings?

Also check lvdisplay settings on the volume since you need to have BAD BLOCK allocation set to NONE to avoid the OS from tromping on the Symmetrix.

What utility are you using to move the data?

[Geoff Wild]

George,

This is the first/only host on the new DMX...

Yes, BAD BLOCK is NONE.

We just did a cp -r -p from old to new....

We maybe o to something - we were also doing an alternate restore test from Netbackup - and we may have clobbered the /sapmnt/XXX dir - even though the restore destination was /XXXtest....

The only other thing we got from HP, was, patch PHCO_27913 - we have 27408, but HP says there are some situations that may cause fil corruption if you don't have the superceded one - PHCO_29913...

Rgds...Geoff

[Geoff Wild]

PEBKAC

Maybe I can get this re-posted under "greatest blunders"

I was testing restore to EMC Parity RAID - and was using tar:

tar -cf - . | (cd /XXXtest; tar -xf -)

Unfortunately, I forgot the cd:

tar -cf - . | tar -xf -

I was in /sapmnt/XXX/exe dir...

DON'T TRY THAT AT HOME!

Rgds...Geoff

[Michael Steele_2]

Is SAP reporting problems or just 'cksum'?

I've gotten new cksum values when new disks and vgs are used to copy the same files from old to new file systems. I believe this is normal.

[Geoff Wild]

Just so you know, this issue is closed.

Michael - cksum and the fact that you could no longer execute the files.

Rgds...Geoff

[Steven E. Protter]

Goeff,

Last Friday I had a bunch of filesystems go bad on a HP-UX box because while I was on vacation my backup filled up the root filesystem. Thought it was good place to stash a 500 megabyte backup. Actually it was a 800 Meg backup but only the first 576 got written before the fs filled up.

Every fs on the box vg00 got corrupt.

This was an 11.11 system L2000 64 bit.

If you had an event where root got full or near full, funny things can happen. Including the symptons you reported.

SEP