[DRBD-user] Default Split Brain Behaviour

Felix Frank ff at mpexnet.de
Mon Jan 24 09:57:38 CET 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi,

On 01/24/2011 01:20 AM, Lew wrote:
> I've encountered some unexpected behavior with a split brain instance.
> It seems from what has occurred that the default behavior is set to roll
> back & discard changes.
> 
> Recently in my sand pit, I've been manually disconnecting resources as a
> an ad hock way of maintaining a snapshot for roll back.
> This way if I'm happy with changes, I can reconnect and within seconds
> we're fully synced again.
> 
> I moved a server yesterday and discovered after a drbdadm connect all,
> that one of the resources had split brained and discarded a few days
> worth of work;
> rolling back to the point in time when the resource was first disconnected.
> 
> What's interesting to me is that the disconnected secondary node had
> never been set primary, so how did we end up in split brain?
> I also do not understand why it was only this resource that split
> brained, when others that existed in seemingly identical configurations
> and states did not.
> 
> I expect I'll need to explicitly prohibit this behavior in a global net
> section covering  after-sb-0pri etc;
> I still don't understand why discard & roll back has been chosen default
> behavior, I'm contending from my experience it should not be.

It's not.

> Looks to me like a few days work is lost, but if anyone knows of a way
> to recover from a roll back discard scenario, I'd be very happy to find out.

Please share pertinent logs and drbd configuration.

Regards,
Felix



More information about the drbd-user mailing list