LVM and disk/device failure

[zoyd_wheeler]

Anyone has had any experience with losing a disk/device under LVM?
I have two PCI controller cards on my fileserver ( Promise ATA cards ).
On these I have built 2 raid-1 arrays using mdadm, both mirrored across the controllers ( one disk from card 1, one from card two ). I have 4 disks -> 2 arrays. On top of this, I run lvm, using one volume group for both arrays, and then on top of this the filsystems on logical volumes.
I should be able to lose a one of the cards without the arrays failing.
However, if one of the arrays needs to resync, the load on the PCI bus becomes so heavy ( since all data will need to be writte across the PCI-bus ) that unless in single-user mode, the machine will hang.
I therefore think of rearraging the arrays, so they are mirrored on each controller, instead of across, and therefore able to resync without loading the PCI bus.
However, in this case, losing a controller would mean losing the array, and therefore - one of the decives the LVM builds on.
Anyone knows if LVM can handle this type of failure withuout losing data? ( Ie automatically stopping / shutting down so that the card can be replaced, and LVM devices back up in a consistent state )

[ectospasm]

If it's really just the controller that's failing, powering down and swapping it with the same model controller should work without LVM even noticing the difference. Of course, if it's the disk/RAID that's failing, you'll need to migrate all of the data off those physical volumes, and then replace the drives. Here's the LVM HOWTO on that subject:

http://tldp.org/HOWTO/LVM-HOWTO/removeadisk.html

Good luck...
_________________
Join the adopt an unanswered post initiative today
Join the EFF!
Join the Drug Policy Alliance!