| Author |
Topic |
|
paulland
Starting Member
7 Posts |
 Posted - 2009-01-13 : 17:34:00
|
| I have a production cluster SQL2K server that I want to upgrade to SQL05. Based on what I know -- not much -- this approach "should" work, and I plan to test it in an offline environment first, but I'd appreciate any feedback on the merits of this approach.I need to maintain the same SQL Server virtual name after the upgrade. And I can't afford a whole new clustered server pair -- but I can afford additional drive space on the SAN.It's my undersanding that a HA cluster (active/passive) can handle multiple SQL Server instances, so long as they're on different disk resources.I'm adding SAN drive space for allocation of "upgrade" partitions for this cluster.I'll install a new SQL05 instance on the cluster using the new upgrade partitions and confirm the new clustered SQL group is operational.When it loooks good I'll shut down the old SQL2K resources group.... move (well, copy) the disk resources to SQL05.... attach the filesystem.... update the dependencies.... reassign the SQL name alias.... and bring up the database on 2005. (Not exactly in that order, I know)Is there anything in general or in particular about this approach that doesn't make sense? And assuming it's a reasonable approach are there any potential hairballs that I should watch for? |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
|
|
paulland
Starting Member
7 Posts |
Posted - 2009-01-13 : 18:24:11
|
quote:
Originally posted by tkizer
You can't just rename a SQL Server so that it uses a new virtual name. You must uninstall SQL Server 2000 and then install SQL Server 2005 with the same name.
I might not have been totally clear on this. I'm not planning to rename the virtual SQL2000 server. I'm taking it out of service as part of the upgrade.I plan to install a new SQL2005 instance on the existing cluster server (alongside the SQL2000 resource), on a separate drive space, with a different virtual name.Writing it out:Current 2K servername\instance\virtual: OLDSERVER\sv1\Data1-- Filesystem on drives D:, E:, F:, G:Proposed 2005 servername\instance\virtual: OLDSERVER\sv2\Data2-- FIlesystem on drives H:, I:, J:, K:Two instances, same cluster, different versions, different SQL resource group, different drive spaces.Then, after verifying the 2005 instance is functional, take down the 2000 SQL resource, mirror the D:-G: data to H:-K:, attach the database to the SQL2005 instance.Finally, rename the 2005 virtual server from \\OLDSERVER\Data2 to \\OLDSERVER\Data1.I'm renaming the 2005 virtual server, not the 2000. The old 2000 server will not be brought back online after the SQL resource is taken down.Does this make more sense? Or is it still unpossible? |
 |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
|
|
paulland
Starting Member
7 Posts |
Posted - 2009-01-13 : 18:29:47
|
[quote]Originally posted by tkizer
This is the part that won't work:[quote]Microsoft says you can do this.. ?http://msdn.microsoft.com/en-us/library/ms178083(SQL.90).aspx |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
|
|
paulland
Starting Member
7 Posts |
Posted - 2009-01-14 : 02:33:43
|
quote:
Originally posted by tkizer
Interesting, I stand corrected. Thanks for the link. This will actually help me greatly in the next 30 days.
So, does my approach seem reasonable? |
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
|
|
dinakar
Master Smack Fu Yak Hacker
2507 Posts |
Posted - 2009-01-14 : 14:19:32
|
| Yes you can have a 2000 and 2005 servers in the same cluster as A/A. Is in-place upgrade not an option here?Dinakar Nethi************************Life is short. Enjoy it.************************http://weblogs.sqlteam.com/dinakar/ |
|
|
|
paulland
Starting Member
7 Posts |
Posted - 2009-01-14 : 14:39:25
|
quote:
Originally posted by dinakar
Yes you can have a 2000 and 2005 servers in the same cluster as A/A. Is in-place upgrade not an option here?
I need to be able to test the process and have an easy rollback path that requires minimal downtime in the event things go awry. The database is substantial (>300GB), and I'd rather not have to restore from backup to get everything back online. An in-place upgrade would definitely be easier, but this is absolutely mission-critical data with typical 24/7 availability, and my prime directive is to minimize downtime. So I want an upgrade process that maximizes the ability to test the end result beforehand, and has an easy rollback path. I'm open to suggestions! |
|
|
|
dinakar
Master Smack Fu Yak Hacker
2507 Posts |
Posted - 2009-01-14 : 16:34:05
|
| Side-by-side can be done in lesser downtime, however lot of planning needs to be in place.. if you have a test environment where you can practice the entire drill, you will feel more comfortable (or not) and that will help you narrow down the choice.Dinakar Nethi************************Life is short. Enjoy it.************************http://weblogs.sqlteam.com/dinakar/ |
|
|
|
|
Upgrading 2K cluster to 05, on the right track?