December 23, 2009 at 3:39 pm
Hello,
We are trying to increase our redundancy and would like to use SQL Server replication to do this. We have about 5 database servers that contain around 1000 databases. Some of these databases are used more often than others. We are a complete Windows shop.
The idea is to create a new big server that can hold all of the databases and have these be slaves of the databases on the 5 other machines. So all data will be replicated over to this new server. If we loose one of the 5 servers our application will immediately switch over and start using the slave version of the database until we can get the other server back up and running. Our application is a mostly read with small updates (occasionally large updates) Does any know if this will work? Are there better ways of accomplishing our goals of redundancy? Are there any links to articles about this?
Thanks,
Pat
December 23, 2009 at 4:00 pm
Thats a big question with no easy answer. Two things I would say though, I wouldn't want to start from where you are and replication would be my last choice for a high availability solution. It would be a nightmare to set up and maintain for 1000 databases.
the closest to your solution of 5 masters and 1 slave server is to set up a mixture of log shipping or full database restores depending on the recovery needs of the individual databases. This will also be labour intensive to set up but once in place easy to maintain and its cheap.
However with that number of databases to achieve redundancy I think you should be looking at either clustering or moving onto VMs.
---------------------------------------------------------------------
December 23, 2009 at 4:17 pm
Thanks for your reponse.
When you say clustering are you referring to a specific product? If so what product is this? With VMs how do you acheive redundancy? You will still need either replcation or backups/restore scripts right?
December 23, 2009 at 4:44 pm
the 'product' is MSCS - Microsoft clustering services, which is at the Os level. You install SQL server on top of that. It allows a physical server to act as a standby for another. google SQL server clustering. As a starter see
http://www.sql-server-performance.com/articles/clustering/clustering_intro_p1.aspx
With a VM (virtual machine) you can cut an image of the server and quickly restore it to another VM.
Both mean a complete change in direction from your current set up and associated start up hardware costs.
Be aware that with SQL 2000 you require enterprise edition to be able to use clustering.
---------------------------------------------------------------------
December 23, 2009 at 6:53 pm
With those servers and thinking DR and switchover, clustering will work, but it requires hardware to make work. I'm not sure if you could do a 6 node cluster, either.
I might suggest that you consider upgrading to SQL 2005 or 2008 and using Database Mirroing instead, which would give you automatic failover. If you don't want to upgrade, or can't, then log shipping might be your best choice.
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply