Cluster 3 nodes

#21

Quote:

Hi,

I tested a failover cluster solution with Windows 2003 and SQL Server 2005
using 2 nodes with optimal result. Now I'm interested to add a third node
and see what happens, but before that I have some questions.

1 - How nodes will work in case of failover?, will use a round robin
schema?, a random schema?, or always node 1 will failover to node 2 and
only
if this is down will failover to node 3.

It is an algoritm the cluster will base its decision on, based on the
following
- possible nodes
- preferred nodes
- AntiAffinityClassNames (not widely used, but used as default for Exchange)
the full description of the algoritm here:
http://msdn.microsoft.com/en-us/library/aa369651.aspx

Quote:

2 - There's any option to create a cluster where only two nodes will
failover automatically and the third manually?, we're thinking on this
option
to have a node remote and only put online if local cluster is unavailable.
In this schema we'll only use the virtual sql name capability of cluster
and
not the failover automatically.

If you put a third node in a seperate location, you are effectively creating
a multi-site cluster.
You must also replicate the data, and some sort of integration software need
to be installed to integrate this replication into your cluster. I would
advice you to contact your Storage Vendor, they will be able to advice you
on this.

Quote:

3 - More generally, I think that when enough resources is better to have
more instances of SQL in the same machine for differents applications than
to
have only one instance with several databases, I'm thinking in HA when one
of
the applications generate that SQL Server use a lot of SQL cycles, a lot
of
memory or the tempdb heavily. What do you think?

the "unit of failover" in a cluster is a "group", which for SQL is an
instance. So if you want your applications to stay up, when only one
instance has a problem and is in process of failing over, then more
instances is better in regards to HA. However, each instance is effectively
running an SQL server engine, therefore, more instances come with an
additional load on your cluster nodes. It is a trade off between HA and
performace. In the worst case scenario you want to be able to run all your
instances on 1 node, if disaster takes out the other nodes in the cluster.

HTH,
Edwin.

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Cluster 3 nodes

microsoft.public.sqlserver.clustering microsoft.public.sqlserver.clustering

Re: Cluster 3 nodes - 09-18-2008 , 04:35 AM