[PVE-User] DRBD confusion in PVE wikki docs.
flavio-pve at zipman.it
Sat Jul 21 13:27:28 CEST 2012
On 07/21/2012 11:15 AM, Muhammad Yousuf Khan wrote:
> so on one side it is written that we need to delete the whole data
> and re-sync and on the other hand it is written that the process will
> take few minutes (probably) because most of the time DRBD updates only
> the changes that have been made since splint brain.
> anyone please help.
I didn't find the places where you copied the sentences from, but I
think you are comparing 2 different situations.
If you have a cluster with a single drbd device and all VMs are running
on a single node, the recovery from a split brain situation is simple,
and it takes few minutes, the time to resync the device.
If you have a cluster with a drbd device and VMs are running on both
node (usually this means you configured LVM on top of drbd) then the
recovery is more complex, because you have correct data on both sides of
In this latter situation you must shutdown all the machine on one of the
nodes, copy all the data of theese VMs on the other node and then
recover from the split brain condition as in the first case above.
At the end you can turn on your VMs. (pay attention: you cannot
absolutely migrate running VMs when you are in a split brain condition
to move them from one node to the other).
To try to minimize recovery time and risk from S.B. it's advised to
create at least 2 drbd device and to take all the VMs on each drbd
toghether on the same node.
A computer is like an air conditioner,
it stops working when you open Windows
More information about the pve-user