SQL Server Performance Forum – Threads Archive

Replacing Server Hardware on Clustered Environment

Dear SQL Gurus, Does anyone know article(s) on the web, or have experience before in replacing server hardware in primary node of SQL Server 2000 Clustered Environment. The great articles wrote by our friend Brad M. McGehee that I found on http://www.sql-server-performance.com/sqlserver2000_clustering_install_part2.asp</a> , discuss the new cluster installation that will install SQL server on both nodes, which is not the case in our environment. I will appreciate very much for any comment or ideas. Have a good day ! [<img src=’/community/emoticons/emotion-2.gif’ alt=’

‘ />] [<img src=’/community/emoticons/emotion-1.gif’ alt=’

‘ />] Cordially, Paulus Tandang Paulus

Hi Paulus, If you have a clustered SQL environment, then you can turn off the node that is not taking the load amke whatever hardware changes you need, then activate it. Similarly the primary node can be failed over to the standby node and then again make the hw changes as you need I don’t quite understand the command about "SQL Server on both nodes, which is not the case in our environment" If you have a SQL cluster then it must be installed on both nodes…? Cheers
Twan

I assume you are talking about an active/passive cluster, where only one node of your server is running SQL Server. Unfortunately, there is no way to remove SQL Server from a single cluster node, they have to be removed from both nodes at the same time. This means your SQL Server will be down during the server replacement. Here is the process, from the big picture view: 1) Backup all of your databases. And document all of your clustering setup so you can duplicate it. 2) Detatch the all of the databases from your SQL Server. I assume that these reside on a shared array, and will stay there during this process. 3) Uninstall SQL Server clustering. 4) Remove Windows clustering. 5) Replace your hardware. 6) Install Windows on the new hardware. 7) Install Windows clustering on both nodes. <img src=’/community/emoticons/emotion-11.gif’ alt=’8)’ /> Install SQL Server clustering on both nodes. 9) Restore the master database from a backup over the current master 10) Re-attach all of the detached databases. —————————– Brad M. McGehee, MVP Webmaster SQL-Server-Performance.Com

Brad & Twan Thanks for your response & great advice, I appreciate it. For Twan: sorry for not being cleared. What we plan in our office is to replace ALL server hardwares on primary node. We have a new server to replace current one, as mentioned by Brad’s response. For Brad: Yes we are talking about active/passive clustering. What I read on the microsoft technet on
http://www.microsoft.com/technet/prodtechnol/sql/2000/maintain/failclus.mspx#XSLTsection129121120120 It mention 3 sections as follow:
– Administering SQL Server Virtual Servers
– SQL Server Setup
– Adding or Removing a Cluster Node from the Virtual Server Definition Those sections did not discuss about replacing the node. What our team plan is as follow for the 1st test:
1.Backup all user databases
2.Failover primary node to secondary node
3.Shut down primary node
4.Bring online new server hardware running on Windows 2000 Advanced Server
5.Run SQLsetup on new server, and setup as virtual server with the same configuration as current server primary node configuration.
6.Failover secondary node to primary node We are thinking this would happen after step 5 above.
The secondary node would make the following attempts to primary node:
– populate all users databases setting without having to run sp_attach_db
– populate the schedule jobs to msdb database
– populate all logins to master database Does any one know any potential risk of corrupting user databases & SQL Cluster installation from taking the above test? If the 1st test above fail, I am thinking about 2nd test by following technet article above by removing current primary node, and add new primary node by maintaining same primary node configuration (virtual server name, instance name, and other required/recommended configuration for primary node) Will the SQL cluster environment recoqnize the new node added as primary node with same configuration as previous primary node that has been shut down, and the two nodes clustered work as before ? I appreciate any comment and ideas. Have a good day! Regards,
Paulus
Paulus

It is important to remove SQL Server, which is not on your list. If you don’t remove it from both nodes, shut down one server, and try to reinstall both nodes as a cluster, it won’t work. It is much easier, simplier, and less stressful, to consider this upgrade as a new install, and in a sense all you are doing is to copy the databases from one server to another, and reattaching them. I have done this twice, and it works well. —————————–
Brad M. McGehee, MVP
Webmaster
SQL-Server-Performance.Com

Hi Friends, Finally we successfully replaced primary node with new server hardware. We are running SQLserver 2000 cluster environment. The process is easy and straighforward as follows:
1. Failover primary node to secondary node
2. shutdown primary node hardware
3. Bring online new hardware on Windows 2000 Advance Server
4. Open cluster admin on secondary node
5. remove primary node
6. go back to new hardware, install Windows Clustering
7. Run SQL Setup and follow instruction to add virtual server to existing cluster configuration
8. Failover back to primary node which is running on new hardware. If the network card is from old server, no configuration needed to recognize SAN drives. If the new server using new card, it has to be configured to map to SAN drives. We use XioTech SAN drives. Last weekend we spend 7 hours trying to failover back to new hardware and did not succeed because we did not realize the new network card has to be mapped to SAN drives using XIOTECH mapping utilities. We decided to use old card that has been configured for the new hardware which make it easier. We installed XML service on the new hardware. When failed back to primary node, the service fail to start and failed over again to secondary node which has no problem on running that service. To correct this problem:
1. On 1st node, open the XML service property for the database that we configure for XML IIS, change the physical path by using UNC path from secondary node to SAN drives.
2. Failed over to 2nd node.
3. Failed over to 1st node, which has UNC path to SAN drives from 2nd drive. \AHLSQLAg

bName, where AHLSQLA is 2nd node physical server name.
4. Take out the UNC path, and set it g

bName
This is what it looks like on the first place on the AHLSQLA under XML IIS for that database property.
5. Failed over back to 2nd node and failed over back to 1st node to test new physical path, and it work with straight network path to SAN drives from 1st node. Hope this helps for everyone ! Paulus

I am going to be replacing my entire cluster with new hardware. It is a 2 node active/active cluster.
Do you have any advice?
Is it possible to keep the virtual server names the same? Thanks

By entire cluster I mean servers and storage sytems. Thnks

It’s possible. Since it’s active/active, Brad’s advice is easier to follow to replace the whole cluster. Paulus

]]>