Each server in the cluster is still an independent server; configuration is
done the same way as a stand alone server. However, you need to configure
with the understanding that both instances MAY have to operate on the same
box.
So if the box has 8 GB of memory, then you don't assign all 8 GBs to the
default instance on that box. If you did, when there is a failover, there
will be no memory for the 'visiting' instance.
If InstanceA has a minimum memory of 4 GB, and Instance B also has a minimum
memory allocation of 4 GB, and the server only has 8 GB, then upon failover,
the 'visiting' instance will be memory starved and will most like not come
online. (You have to make sure that there is enough memory for the OS as
well.)
Let's say both boxes have 8 GB. InstanceA contains large, heavily used
databases, so it is allocated 5 GB as a maximum, and InstanceB is less
demanding, so it is allocated 2 GB as a maximum. Upon failover, the two
instances require 7 GB, and leave 1 GB for the OS and memory management.
'Bean counters' will complain since it appears that InstanceB only needs 3
GB for the server, and adding an addition 5 GB is seen as a waste. However,
you have to consider the cost of the 'unused' 5 GB on InstanceB as insurance
payments for future availability. (And the same for the 'unused' 2 GB on
InstanceA.)
Thing to remember with configuration, is that upon failover, all instances
must be able to work (albeit perhaps not performing quite as well) on one
box. Most Server wide configuration settings 'should' be the same. Memory is
the most common setting that is different between Instances.
--
Arnie Rowland, Ph.D.
Westwood Consulting, Inc
Most good judgment comes from experience.
Most experience comes from bad judgment.
- Anonymous
You can't help someone get up a hill without getting a little closer to the
top yourself.
- H. Norman Schwarzkopf
[quoted text, click to view] "burt_king" <burt_king@yahoo.com> wrote in message
news:91DFAA5E-B271-4CA4-B6FE-727CA857BC39@microsoft.com...
> Roland, very clear, thank you.
>
> Now, slightly off but still for clustering, how do I administer the
> traditional sp_configure settings? I'm speaking of memory for myself but
> that raised the larger question, How do I administer the server? I'd
> assume
> through the virtual server name since on the back end the Master and MSDB
> are
> really being written to by the Active node... Am I correct? If I want to
> give the server more or less memory, can I do so through the Enterprise
> manager interface?
>
>
> "Arnie Rowland" wrote:
>
>> Each instance requires it own dedicated disk resources that cannot be
>> shared, and databases CANNOT be shared between instances. Those dedicated
>> resources cannot have the same drive assignment.
>>
>> Instance A has SAN drive G, H,
>> Instance B has SAN drives M, N
>>
>> Upon failover of Instance A, Physical computer B now runs both Instance B
>> AND Instance A, and Drives G, H, M, N.
>>
>> This has been a very simplistic example, but hopefully you now understand
>> a
>> bit better.
>>
>> --
>> Arnie Rowland, Ph.D.
>> Westwood Consulting, Inc
>>
>> Most good judgment comes from experience.
>> Most experience comes from bad judgment.
>> - Anonymous
>>
>> You can't help someone get up a hill without getting a little closer to
>> the
>> top yourself.
>> - H. Norman Schwarzkopf
>>
>>
>> "burt_king" <burt_king@yahoo.com> wrote in message
>> news:C926C991-9D66-4531-A5FF-912AEC097165@microsoft.com...
>> > My employer is asking to do an Active / Active SQL 2000 cluster to
>> > utilize
>> > both servers. Problem is, I don't think SQL 2000 supports an
>> > Active/Active
>> > solution because both would need access to the disk which isn't
>> > possible...
>> > The only time an Active / Active cluster can work is when you have
>> > multiple
>> > instances which allows Node A to use the disk for one instance and Node
>> > B
>> > to
>> > utilize the disk for a different named instance.
>> >
>> > Can someone clarify for me? Thanks.
>> > --
>>
>>
>>