[PVE-User] DRBD confusion in PVE wikki docs.

THe_ZiPMaN flavio-pve at zipman.it
Sun Jul 22 00:34:04 CEST 2012

On 07/21/2012 07:37 PM, Muhammad Yousuf Khan wrote:
> here is the proxmox official wikki doc.
> http://pve.proxmox.com/wiki/DRBD
> i agree with Robert, what document saying is ,
> it suggest us to create two volume r0 and r1
> on node1 we use r0
> and on node2 use r1
> if split brain occurred then we dont have to shift whole data we just
> discard the old data from victim node and resync the drbd data.
> for instance in my confusion case if we discard the data from node1 r1
> and node2 r0
> in case of split brain. so the concept of saying it will take few
> minutes as it updates only the changes, doesn't fit in the scenario
> and making me confuse.

If you fall in a split brain condition with drbd but on one node there
were no VMs running, the data on that drbd device will be intact and
will be a "snapshot" in the instant when the SB occurred (noone wrote on
it after that moment).

In this case you don't need to scratch the content of the volumes,
because the data is consistent. You simply reconnect the device to it's
primary and it will resync copying the data written after the S.B. occurred.

Instead in the case where you have only one drbd device and 2 node that
write on it, both the copy of the device contains fresh data and so you
must do a manual merge. After that a full resync (or better the
re-creation of the drbd device) must take place.

Flavio Visentin

A computer is like an air conditioner,
it stops working when you open Windows

More information about the pve-user mailing list