topic about bad block =off in Operating System - HP-UX

about bad block =off

rayche — Tue, 28 Feb 2006 20:13:37 GMT

Hi everyone,

HP said in root,swap and dump devce, the option bad block should be off, I think there is problem exist, if system read a bad block of root device, because bad block=off, so system will not mark the bad block ,and try to read the block continually,this will consume all CPU time and system has no any other CPU time to handle the request come from client! although we have mirror disk, system will not swith to another system disk!!! I think this is a big problem. how can us sovle it ?

thanks

Re: about bad block =off

rayche — Wed, 01 Mar 2006 02:26:31 GMT

HI,

who can tell me how to understand it ?

Thanks.

Rayche

Re: about bad block =off

Tiziano Contorno _ — Wed, 01 Mar 2006 05:14:18 GMT

Interesting question.

Actually "the system will mark the block BAD, and will not try to read it again".

Bad block relocation policy has two switch "n,N":

[n]: Prevent bad block relocation.
Upon media failure, LVM will mark the failed block as bad in the Bad Block Directory, but will NOT attempt to relocate the bad block to a new location on disk. Future I/O requests which contain the bad block will return with an I/O error. No attempt will be made to access the bad block.

[N]: Disable bad block relocation and the Bad Block Directory.
Upon a media failure, LVM will NOT attempt to relocate the bad block. In addition it will NOT enter the block in the Bad Block Directory. LVM will have no record of the block being bad, and will attempt to access it on future I/O requests.

Maybe this is intended to let you know if critical LVs have bad blocks.

As for mirror copy switching it is triggered by a failed PV.

Re: about bad block =off

rayche — Wed, 01 Mar 2006 20:31:37 GMT

thanks for Vanadio's help. In fact,I met the problem last time, I used RP8400 machine, and at that time, when I try to telnet to the machine, there was no response, so I went to DC room to check using console, I found it was very slowly when I typed command, then I use sar command to check,I found disk wio is 100%, I used mstm command to find there was bad block error. I do not understand why system try to read a bad block error ? in that system, root,swap and dump's bad block option is off, other is on.

Thanks.