[linux-lvm] Curious about LVM PV failure and mitigation

Ben Kingston benjamin.kingston at bencloud.net
Fri Sep 18 01:22:44 UTC 2009


I've been wondering for a while what happens if a drive fails that is
part of a VGroup (spanned across PV's, not Raid0), specifically in a
active VG. Does LVM continue working as though there was a large sector
failure? What happens if the first physical drive that the VG was
created on fails? Is the entire Volume lost?

I've hypothosized a mitigation method for the first physical drive
failing. I would create the VG on an SD card so that all of the metadata
would be stored there, and if a moving drive fails, it wouldn't take the
LVM with it. This would also an easy way to have a remote dd image of
any other important volume metadata (e.g. Luks). But before I do this,
I'd like some ideas on paramaters. Would turning on the write protect
lock cause any issues, after all PV's are added, to writing to the HDD
with EXT4? What would be an optimal SD size? I'm guessing 256MB would be
sufficient, but I don't want to worry about messing up EXT4's superblocks.

Thanks for the help.




More information about the linux-lvm mailing list