There is a trace flag that will disable the disk dependency check, but why
would you do that?

The whole point of a failover cluster to provide multiple hardware
components to support a virtual interface. These items are placed in a
resource group such that if any one of them fails, the whole group fails,
and is then picked up and presented on an active member node.

If you had multiple instances running in the same resource group, not only
would they have to share the same disk, but the same IP address and network
name as well. If one were to fail, then they all would fail. You want to
limit your failure points to just a single instance, but affect all
resources required to maintain that instance.

Keep this in mind: when a failure occurs, there is still a service outage.
It is just automatically detected and recovered, but it is still an outage
(much shorter than on a single host) but an outage nevertheless. Why would
you want the availability of one instance to be affected by another? If
that is acceptable, consolidate, don't multi-instance. Host all of your
databases on the same instance. This way, they all fail together, but in a
supported configuration.

Sincerely,


Anthony Thomas


--

[quoted text, click to view]

[quoted text, click to view]

If you cannot create a additional group, due to lack of seperate disk
resources or other reasons, then I would agree with Anthony here. Host all
your DB's in the same instance.

rgds,
Edwin.