Video Screencast Help

Upgrading SFWHA v5.1GA to SFWHA v5.1 SP1

Created: 18 May 2011 • Updated: 01 Jun 2011 | 1 comment
This issue has been solved. See solution.

Currently I have a 6 node cluster (5 active, 1 passive) running SQL 2003. I would like to perform a rolling upgrade of all the nodes clusters from SFWHA v5.1 GA to SFWHA v5.1 SP1 without taking any downtime on any of the active cluster nodes.

My question is can a rolling upgrade be done when upgrading from SFWHA v5.1 GA to SFWHA v5.1 SP1? Do I have to completely shut down the entire cluster? Or can I upgrade the passive node first, then failover one of the active nodes to the passive node and then upgrade that node (original active node).

Thanks, Derek

Comments 1 CommentJump to latest comment

rhanley's picture

Hi Derek,

      I would strongly recommend upgrading to the latest version which is 5.1 SP2. The release notes included in this download provide instructions for performing a rolling upgrade in the Installation and Upgrade Guide. Unfortunately, performing a rolling upgrade still requires moving groups between the active and passive node(s) during the upgrade process, so there will still be required downtime as these applications fail back and forth between nodes.

You can download the Installation Guide for 5.1 SP2 (pdf) from the following location:

Page 142 begins the discussion on Rolling Upgrade (listing an overview of the tasks and limitations), and this is followed by the detailed procedure on performing the rolling upgrade.

The instructions provided assume a 2-node cluster (Node1 and Node2) with Node1 being used to represent your 'Active' cluster nodes, and Node2 representing your 'Passive' cluster nodes.

To Summarize, you will be upgrading your Passive Nodes first, then moving your Application groups to these servers. Once online, you will then upgrade the original active nodes to complete the upgrade operation. Moving these groups from the Active to the Passive nodes is the only required downtime for the operation. There will be times when the HAD engine is stopped so it is not monitoring your application, but the application will remain online during these times.

Since you have 5 active and 1 passive, you will want to upgrade as follows:

Assuming Nodes 1-5 are active and Node 6 is passive

- Upgrade Node 6 (Passive)

- Move Groups from Node 1 to Node 6
- Upgrade Node 1
- Move Groups from Node 6 back to Node 1

- Move Groups from Node 2 to Node 6
- Upgrade Node 2
- Move Groups from Node 6 back to Node 2

- Move Groups from Node 3 to Node 6
- Upgrade Node 3
- Move Groups from Node 6 back to Node 2

Since you are dealing with a single passive node, your process will require a short outage for each cluster group in order to move it to the passive node, then back to the active node following the upgrade. Unfortunately, this is the best option for you in limiting the amount of downtime required.

This process assumes you are knowledgeable with HA clusters and I'd strongly recommend you reading through all of the steps and ensuring you understand each of the operations.

If you have a lab cluster where the process could be attempted on first, that would be ideal. Otherwise, if you are unclear on any of the information, I would recommend posting your concerns here, or a better option may be to open a Support case and allow a TSE to assist you in your understanding of each of the steps so you can feel confident in the process.

I hope this helps,