May 5, 2015 at 6:03 am
I am currently in charge of a SQL server. This is a single server with 9 named instances using two node Failover Clustering. I have just be charge with upgrading the servers from SP2 to SP3. What is the correct process for this. Can I update the passive server first during the work week and then update the active servers over the weekend. Can I update once instance at a time, failover and then update the corresponding instance on the other server? Any other issues that might come up. Thank you
May 5, 2015 at 6:18 am
If I understand correctly, both nodes are currently active for some instance.
In this case, upgrade the first node selecting just the instances that are not running on the node. Do the same on the other node.
When finished, fail over the instances to the other node (already patched: this upgrades the instance). Do this one instance at a time to keep the load on the nodes balanced.
Now you just have to apply the service pack to the passive nodes for the instances you already have upgraded.
-- Gianluca Sartori
May 5, 2015 at 6:59 am
Thanks for the reply. this is my first time working with clustered SQL server so I am a little lite on the terms. By active do you mean that both SQLDE on each node/instance is running? This is not the case.
ie
node 1 SQL1 SQLSDE running node 2 SQL1 SQLSDE not running
I been tasked to update node 2/all instances during the work week and then node 1/all instance over the weekend. Do you need any other information?
May 5, 2015 at 7:05 am
So, 2 nodes, multiple instances, all running on one node.
Run the service pack on the passive node (the one with no instances running), move all intances to it when done, update the first node.
-- Gianluca Sartori
May 5, 2015 at 7:09 am
That is exactly the set up, and that is what I thought. Would I be able to update the passive instances during the workweek and then on the weekend fail over and complete the job?
May 5, 2015 at 7:18 am
It's a good strategy. Go for it.
-- Gianluca Sartori
May 5, 2015 at 7:23 am
That scenario has worked well for me in the past. It also has the benefit of getting half of the work done during the week. 🙂
Please be sure to back everything up before you start. You don't want to be caught wishing you had a full backup and not having it. You know...just in case. Then again, when it comes to upgrading, I admit I'm more than a little bit paranoid. Also, you did practice in a test environment with some sample databases from production, right?
May 5, 2015 at 7:30 am
In a prefect world yes I would test it first in our test environment, except for the fact that they don't have a test environment. As for the backup I am equally as paranoid. I don't run any updates SQL or third party without taking a full SQL backups (system and user) and getting a VM snapshot of the server(s).
Thanks for all of your replies
And one more question
Can I update each instance at different times? Failover one instance and then update that or do I have to do all at once?
May 5, 2015 at 7:36 am
The real upgrade happens when you fail over the instance: installing the service pack binaries doesn't affect the availability of your instances, so I don't think there's any need to update them one by one.
If you're asking whether it's a good idea to fail over the instances one by one, then yes. If something goes wrong during the execution of the upgrade script, you want to know for the first instance so that you can prevent the same failure from happening on the other instances.
-- Gianluca Sartori
May 5, 2015 at 7:36 am
Joshua D Gang (5/5/2015)
In a prefect world yes I would test it first in our test environment, except for the fact that they don't have a test environment. As for the backup I am equally as paranoid. I don't run any updates SQL or third party without taking a full SQL backups (system and user) and getting a VM snapshot of the server(s).
Oh my. I have to bring up SQL 2000 SP3, which was almost immediately replaced with SP3a because SP3 was a disaster.
I found the same thing when I tried to patch SQL 2005 from SP3 to SP4 and SP4 rendered my server a paperweight. Thank goodness that was in test.
Finally, the recent SQL 2014 SP1 was pulled almost immediately because it was taking down servers.
I would really try to get a test server, even if only for a little while, to test the installation. Tell them they can have it back when you're done. I know there's time and money involved, but it probably won't compare to the cost of down time.
May 5, 2015 at 7:41 am
Thanks for all of your help. We will see how it goes. I am meeting with the CIO and SAN/system admin today and see about getting a test servers
May 5, 2015 at 9:24 am
Joshua D Gang (5/5/2015)
Thanks for all of your help. We will see how it goes. I am meeting with the CIO and SAN/system admin today and see about getting a test servers
Management may just see time and dollar signs, but if you contrast the cost (in time and money) of testing with the cost of down time, they might decide that the risk isn't worth it. Good luck in your meeting. I hope it goes well. If not, at least you have the CIO's buy-in on the risk.
Viewing 12 posts - 1 through 11 (of 11 total)
You must be logged in to reply to this topic. Login to reply