Newbie in Clustering

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Pasquale - 01 Jul 2008 16:57 GMT

I have two servers. On the first server I want to put two dbs and on to the
second second server I want to put the third db.
The dbs on the first server will be queried from 600 users. The high
performance is an important goal, and the high data availability, also.
I think to install two SQL Server instances on the first server and other
two SQL Server instances on the second server.
On the first SQL instance of the first server I want to put the first two
dbs and on the second SQL instance the third db. In the same manner, I want
to use the SQL instances in the second server. Is it correct?
Now, do I create a cluster with two nodes or two cluster with two nodes, in
respect with performance and availability?
Thanks

Reply to this Message

Linchi Shea - 01 Jul 2008 17:26 GMT

You have three databases, and for some reason you want to place two of them
in one instance and the 3rd one in another instance. So you need two
instances in total, but you seem to be suggesting four instances. There is no
need for the other two instances.

If I were you, I'd first try the simplest option: one SQL instance in a
two-node cluster, and put all three databases in that instance.

If there is a strong reason you can't have all three databases sharing an
instance, I'd create two SQL instances in a two-node cluster.

If there is a performance problem with the databases and/or instances
sharing a server, have a two-node cluster for each instance.

Linchi

> I have two servers. On the first server I want to put two dbs and on to the
> second second server I want to put the third db.
[quoted text clipped - 8 lines]
> respect with performance and availability?
> Thanks

Reply to this Message

Pasquale - 01 Jul 2008 17:47 GMT

I consider the simplest option: for example, I install one SQL instance on
the first server and no one on the second server. The first server is the
first node and the second server is the second node for the unique cluster;
the SQL instance contains all three dbs, isn'it?

Now, I try to consider a performance problem. The first two dbs will be used
from more 600 cuncurrent users. In order to implement a two-node cluster for
each instance, I must one SQL instance on the first server and one SQL
instance on the second server. Each instance has all three dbs. The first
cluster has the first server as the first node and the second server as the
second node; the second cluster has the second server as the first node and
the first server as the second node, isn'it?

Many thanks

> You have three databases, and for some reason you want to place two of them
> in one instance and the 3rd one in another instance. So you need two
[quoted text clipped - 24 lines]
> > respect with performance and availability?
> > Thanks

Reply to this Message

Geoff N. Hiten - 01 Jul 2008 18:38 GMT

I think you misunderstand. You do not install an instance on a node. You
install the instance to a cluster. You can choose which node an instance
runs on right now, but instances must be able to failover to another node to
maintain high availability. If you have two nodes and two instances, be
sure any one node can support all the instances at once, otherwise you don't
have a highly available solution. You can run odd numbers in a cluster
where you have two instances and three nodes. You can intentionally "stack"
instances on nodes of you want, you just have to make some decisions about
memory and CPU resource allocation.

Signature

Geoff N. Hiten
Principal SQL Infrastructure Consultant
Microsoft SQL Server MVP

>I consider the simplest option: for example, I install one SQL instance on
> the first server and no one on the second server. The first server is the
[quoted text clipped - 51 lines]
>> > respect with performance and availability?
>> > Thanks

Reply to this Message

Pasquale - 19 Jul 2008 15:49 GMT

The goal is to use the two available server (my cluster nodes) and to
distribute the dbs on two different SQL Server instance.
In order to do it I think to use a multiple instance cluster (a/a) with two
nodes, Node1 (or Server1) and Node2 (or Server2).
I will execute the install of an instance on an initial node (for example
Node1) and then I will decide which the node runs the instance installed and
then I will setup the second instance that shall run on the other node (in
this case Node2), isn't it?

Thanks

> I think you misunderstand. You do not install an instance on a node. You
> install the instance to a cluster. You can choose which node an instance
[quoted text clipped - 61 lines]
> >> > respect with performance and availability?
> >> > Thanks

Reply to this Message

Geoff N. Hiten - 20 Jul 2008 03:51 GMT

Pretty close. Be sure to handle memory allocation between the instances
during a failover so that one instance is not starved.

Signature

Geoff N. Hiten
Principal SQL Infrastructure Consultant
Microsoft SQL Server MVP

> The goal is to use the two available server (my cluster nodes) and to
> distribute the dbs on two different SQL Server instance.
[quoted text clipped - 90 lines]
>> >> > respect with performance and availability?
>> >> > Thanks

Reply to this Message