Uncorrectable read error on volume

[Miguel Ward]

Hi:

We are running Tru64 4.0D with LSM enabled. Have lost the mirroring on the root volume and when trying to rebuild with volrootmir -a rz9 I get an uncorrectable read error on the volume from which I want to copy which aborts the whole operation.

I know the block number (12818) with the error but don't know how to work round this.

First of all does 'volrootmir' copy physically block by block?, if this is the case can I mark this block as unreadable or some other alternative? What command should I use.

If the copy is logical how can I identify the file which belongs to that position (I am using advfms)? which I would then try to erase/overwrite.

The computer itself plus all the critical software is running ok (including normal reboots) so it would seem that the damaged block is not critical.

Another option, perhaps, is to tell volrootmir to ignore this error but can't seem to find option to do this.

Screen output of above command:

Mirroring rootvol to rz9a.
root/volplex: Volume rootvol, plex rootvol-01, block 12818: Plex read:
Error: Read failure

I am an Oracle/OpenVms man so pretty lost, all help appreciated.

Thanks

Miguel

[Hein van den Heuvel]

>> Have lost the mirroring on the root volume

Did you loose that before this failure, or because of the failure?
You are not trying to force the system to rebuild the wrong way 'round are you?

If rootvol = 01 + 02, and 01 is deemed broken then it should be replaced and a new plex build from 02.

If this was happening to me, I would start by taking copies of all volatily structures on the root volume, notably /etc.
If one block is bad, others may go soon.

If I had nothing left to loose, then I might try to force a write to the back blocks. The underlying scsi may then revector to good blocks. You would need to talk to the raw device. Try reads first. For the reported block, and some probing around (behind) it. dd -iseek ...

When you found the bad zone use dd -oseek to write nulls (or anything else really) to it.

Actually.... do you still have the other plex? Even though as a whole that would be stale, the bad block might be better than null's :-). Try dd to copy that zone from a stale plex to the broken plex and somethign good might come out of the two bad things.

Ask professional help!? HP Support?
Data recovery services?



>> I am an Oracle/OpenVms man

Nothing wrong with that!
On the contrary
:-)

Hope this helps some,
Hein van den Heuvel (at gmail dot com)
HvdH Performance Consulting

[Hein van den Heuvel · Accepted solution]

Update?
Hein.

[Miguel Ward]

Hein:

Thanks for assistance and sorry for delay.

I was hoping there was someway to simply mark the block as invalid or identify exact file to erase/recover from backup. I have never used dd command and considering risks involved and how critical system is I will be requesting help from original provider of software, etc.

Miguel