June 1, 2005 at 4:08 am
If I set up an active/active SQL Server using 2 servers at 2 different geographic sites, is it possible to connect application users to only the server on their site? I am not sure this is possible considering that the 2 sql servers in the cluster share a single name and IP address. I just want to know if anyone has done this before.
The reason for this, is to have 2 sql servers on 2 separate sites share a data drive which a cluster provides. Replication is not an option because the link between the 2 sites is very slow and users are complaining. In addition a bi-directional replication will not resolve the issue as the application users can only connect to one server at a time.
Regards
June 1, 2005 at 6:41 am
I guess that you don't understand something about clusters. If you set up active\active cluster you will have two virtual SQL servers, two IP adresses and two network names and they must be different. It is not possible to have only one network name and one IP adress in active\active configuration
Hope this helps
June 1, 2005 at 9:08 am
Thanks for pointing out this. I remember now that there will be one default and one instance databases in an active/active cluster. In that case each site application can be connected to one of the virtual servers while still sharing the shared data drive. I suppose if any of the 2 virtual server fails, the cluster will fail over to the other one without the users being aware of this. Am I correct on this.?
June 1, 2005 at 9:47 am
The data drive is shared but is dedicated to the instance. The data drive is not shared between instances.
When a cluster fails over, you will witness a brief outage. SQL Services gets re-started when the cluster fails from one node to the second node.
June 2, 2005 at 3:32 am
Thanks. But I do not understand what you mean by the statement below:
The data drive is shared but is dedicated to the instance. The data drive is not shared between instances.
In an active/active setup, I expect the 2 instances to be able to access the data drive and make changes as necessary. If not what makes the arrangement active/active.?
Why can't users' application be connected to the two instances and make changes to the data on the shared drive?
June 2, 2005 at 11:46 am
If you will use MSCS (Microsoft Cluster Services), you cannot use a slow connection, whem you use a MCSC a latency packet network cannot be < 500ms.
With Veritas Cluster, you will have a same limitation.
If a Network is slow whem you done a data replication ?
June 2, 2005 at 12:27 pm
Microsoft clustering is basically a 'shared nothing' implementation. That means that you don't have load balancing between nodes/virtual servers. One cluster node will own the disks and users have to connect to the virtual SQL Server running on that node. If the instance fails over to the other physical nodes the disks associated with that virtual server go to the other physical node and are started up.
Active/Active clustering just means that you have at least two virtual servers running on the same cluster. They could be owner by any of the physical nodes in the cluster. All this stuff gets even more confusing in up to four node clustering in Windows Server 2003 (Enterprise Edition, or whatever they are calling the most robust version of Win2003).
Check out clustering in Books Online for more info.
June 3, 2005 at 8:54 am
Thanks ChrisMoix. The ownership issue is the key.
Replication cannot be an option because there are 8 databases and the link is slow.
Thanks to everyone for the insight provided. I now the limitations of clustering better.
Viewing 8 posts - 1 through 7 (of 7 total)
You must be logged in to reply to this topic. Login to reply