What's wrong with bad block replacement

[Richard Munn]

I have just fixed a system which got a bad block in it's swap (lvol2) logical volume. The LV was mirrored but refused to resync because one of the PE's had a bad block in it.

What I did was use lvchange -r to turn on bad block relocation then an lvsync and everything is normal again.

But in reviewing the LVM documentation it explicitly says NOT to turn on bad block relocation for the boot, root and dump - does not mention swap but other people have told me not to set it for swap either. It also says that by default bad block relocation is turned on. This is wrong, by default it turned off. But my real question is why is bad block replacement so bad for boot/root/dump (swap?)? Surely this means that a single media error will take out the whole system! A local engineer said that if you get a bad block error the whole disk needs to be replaced. This sounds a bit dramatic especially when other UNIX's seem to handle bad blocks quite transparently.

[avsrini]

Hi Richard,
Check the following

http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000063198322

As per the doc, Bad block relocation is not supported by hpux for root / boot / dump and primary swap. But why it is not supported, only hp engineers has to explain us. You raised a very good point here. Why we have to replace the whole root disk if there is a single bad block occurs?. I'll try to get more info on this, if i get I'll post here.

Srini.

[Steven E. Protter]

I you ever get into a problem with disk and timeouts, you don't want bad block relocation on.

If you get a bad block, you don't want that disk on your HP-UX system any more.

While your system is still running, make a make_tape_recovery backup, arrange for hardware to replace the disk and save yourself some pain.

Ignore HP's advice on the OS at your peril.

try the dmesg command. If you get an lbolt, let me shout....GET RID OF THE DISK

Thank you very much.. That felt good.

Steve