Two Datacenter/Two nodes SG cluster/Two Mirrored Arrays

[Pedro Cirne]

Hi,

In a 2 datacenters, 2 nodes SG cluster, 2 host-based-mirrored arrays, what's the best option for the cluster lock?

Thks!

[melvyn burnard]

you have the choice of a Dual Cluster Lock, or use a Quorum Server
I would go the QS route personally

[Pedro Cirne]

Thks Melvyn, I'm not sure how things will work in the following cenarios:

With the QS, if the QS is located in the datacenter where the disaster happens, will the node in the surviver datacenter halt or not? I think it will and this is a problem...

With the dual lock, if LAN connectivity is lost between the datacenters, what will happen? Which of the nodes will survive?

Would you clarify on this?

Thankyou very much!

Pedro

[melvyn burnard · Accepted solution]

>With the QS, if the QS is located in the datacenter where the disaster happens, will the node in the surviver datacenter halt or not?
Yes, the remaining nodes will TOC, but you could restart the cluster afterwards with just teh surviving node


>With the dual lock, if LAN connectivity is lost between the datacenters, what will happen?
You end up with the probability of having split brain syndrome, where each side will attemt to continue as part of the cluster, and this will probably result in data corruption.
Ideally the QS should be in a 3rd location, and probably better in the DR side.

[Pedro Cirne]

I completely agree with you, I was wondering if there was another option, "bullet proof" but there isn't :)

A good improvement with SG would be to have a lock mechanism based on a shared lvol mirrored on both arrays.