device-mapper and LVM on DRBD

D

DRVTiny

New Member
Feb 5, 2010
27
0
1
I have two nodes with /dev/drbd0 as a physical volume and volume group "replicated" on it.
When i create or rename logical volume in vg "replicated" on one node, it's appear on another, but only in the output of lvdisplay/lvscan/lvs.
/dev/replicated/ and /dev/mapper stays unchanged and live migration with such logical volumes not working. How to enforce device-mapper to reread logical volumes in volumegroup and recreate device nodes in /dev/mapper and symlinks in /dev/replicated?
vgchange -a n replicated && vgchange -a y replicated works for me, but it is too destructive...

PLEASE, help me. i need to recover my volume group after modifying metadata with standard lvm utilities, not with proxmox lvm metadata locking system (which is similar to clvm) :(
 
Last edited:
Because i need to stop all virtual machines running on the target node or live-migrate them everytime i made changes to the LVM manually.
You may have a question: why i need to do any changes manually? Because Proxmox cant resize virtual machine images on the LVM, but lvresize does it!
 
sorry, but I do not understand why 'vgchange -a y' is destructive?
 
Because it cause downtime of the services and downtimes is definitely destructive for bussiness!
 
dietmar said:
Sigh, seems I do not understand you. Running 'vgchange -a y' does not cause any downtime here.
Click to expand...
To activate volume group ("vgchange -a y") i need to deactivate it first ("vgchange -a n"), but i cant do it without halting all virtual machines and maybe all Proxmox services too!
 
I did recover my LVM and DRBD, but again i dont know how. One recomendation is to disable "primary-on-both" temporary and try to recover as it was Primary/Secondary split-brain, not Primary/Primary.
 
You must log in or register to reply here.
Top Bottom