March 14, 2011 at 9:19 am
I currently have a single SQL Server 2005 Standard 64-bit server running on Windows Server 2003 Enterprise. To plan for redundancy I'm looking at using the Microsoft Failover Cluster Services. To do this, I need to purchase a second server, but I have a few questions.
1. Can I use different hardware for both nodes? If so, how similar or different do they have to be? The problem is, our current DB server is a HP Proliant DL585 G2, and HP's currently selling G7 models. So while we can match the memory and local drive sizes and the attached storage will be identical, we'd still have 4 processors (8 cores) on the old server and 4 processors (48 cores) on the new server. In the SQL Server properties I've set it to automatically detect processors, will this cause problems? Or as we fail back and forth will it automatically figure out it has more or less processors.
2. Is it possible to create a 1 node cluster, then add the second node later? The reason is, I have a small amount of downtime to create the new setup. If I can configure the new server as a single node cluster, then join the old one after we've migrated everything over this would make it easier.
3. On the existing server, will I need to uninstall the existing SQL Server to and let the cluster manager install SQL Server for me, or can I just leave SQL Server installed and tell the cluster to manage the existing server?
March 14, 2011 at 9:32 am
1. No, you should have identical hardware for each node. I can't remember whether this is compulsory or just highly recommended. You should also check that the hardware you do have is on Microsoft's approved list for clustered servers.
2. Yes, you can create a single node cluster.
3. You need to uninstall and then install SQL Server as a clustered instance.
John
March 14, 2011 at 9:34 am
mattboston (3/14/2011)
I currently have a single SQL Server 2005 Standard 64-bit server running on Windows Server 2003 Enterprise. To plan for redundancy I'm looking at using the Microsoft Failover Cluster Services. To do this, I need to purchase a second server, but I have a few questions.1. Can I use different hardware for both nodes? If so, how similar or different do they have to be? The problem is, our current DB server is a HP Proliant DL585 G2, and HP's currently selling G7 models. So while we can match the memory and local drive sizes and the attached storage will be identical, we'd still have 4 processors (8 cores) on the old server and 4 processors (48 cores) on the new server. In the SQL Server properties I've set it to automatically detect processors, will this cause problems? Or as we fail back and forth will it automatically figure out it has more or less processors.
2. Is it possible to create a 1 node cluster, then add the second node later? The reason is, I have a small amount of downtime to create the new setup. If I can configure the new server as a single node cluster, then join the old one after we've migrated everything over this would make it easier.
3. On the existing server, will I need to uninstall the existing SQL Server to and let the cluster manager install SQL Server for me, or can I just leave SQL Server installed and tell the cluster to manage the existing server?
First of all, the hardware needs to be on the cluster hardware compatibility list.
1a. Ideally, you should use identical hardware for each node.
1b. Remember that when SQL "fails over", it actually completely restarts on the new node. So it should figure out how many processors it has. Having different hardware means that SQL will run differently on each node of the cluster.
2. Yes. This is actually one way where you can upgrade a server without having to change the application's connection info. (Note that this means that the application(s) will need to be changed to the cluster name when you set up your cluster.)
3. Yes. You need to run the "Install SQL on a cluster". It is separate from a normal install. The first install establishes the cluster instance, the install on the second node "just" joins to the cluster.
Wayne
Microsoft Certified Master: SQL Server 2008
Author - SQL Server T-SQL Recipes
March 14, 2011 at 9:48 am
WayneS (3/14/2011)
First of all, the hardware needs to be on the cluster hardware compatibility list.
1a. Ideally, you should use identical hardware for each node.
1b. Remember that when SQL "fails over", it actually completely restarts on the new node. So it should figure out how many processors it has. Having different hardware means that SQL will run differently on each node of the cluster.
If we used different hardware, during normal operation I'd want the new server to be the active node, so we'd only be using the old server when we're either doing maintenance on the newer server or the newer server is down.
If both servers are on the compatibility list, are there any other reasons why it shouldn't be done?
March 25, 2011 at 9:06 am
buy the DL 380 models unless you have some huge VLDB's. the 500's series have better CPU's and cost way too much since they are meant for the largest companies
for the price of a 585 or 580 G7 you should be able to buy 2 DL 360 G7's easily
Viewing 5 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply