June 28, 2006 at 8:11 am
Hi Everyone,
I have a 3 node cluster group with Active/Active/Active configuration. Let us assume that node names are Server1, Server2, and Server3. Replication is also implemented on this 3 node cluster group. Server1 should work as a Publisher and also as a distributor. Server2 and Server 3 should work as subscribers to Server1 on normal conditions. But, If Server1 fails by any reason then one of the subscriber (either Server2 / Server3 or both Server2 and Server3) should take the Server1 responsbility as a publisher and continue the database process as normal conditions. I have few questions on this 3 node configuration and replication set-up which are mentioned below.
1) I don't know how a 3 node Active/Active/Active cluster configuration set-up and working process.
2) How to configure a subscriber or both subscibers (Server2 / Server3) as publisher in case of publisher (Server1) fails by any means.
I greatly appreciate your time and help. If you have any questions please feel free to contact me at bhushan_kalla@yahoo.com
Thanks,
Bhushan
June 28, 2006 at 8:47 am
Bhushan
The point of clustering is if a resource (for example SQL Server) fails on one node, it stops and continues running on another node. Therefore you shouldn't find yourself in a position where Server1 isn't available, unless it's due to database corruption or multiple disk failure.
Replication isn't designed to be dynamically switchable in the way you're asking. When you set up a publication on a database, you alter the schema of that database, and so if you wanted a subscriber to take over as a publisher you'd have to set up replication from scratch (and configure the other subscriber to subscribe to a different publisher). You could script it all out and then spend a lot of time and effort testing all eventualities. Far better, I think, to devote your efforts to a backup and recovery strategy.
I've never worked with a three-node cluster, but even if I had, I don't think I would be able to tell you how it all works here. Have a look on the Microsft website for white papers on clustering.
John
June 28, 2006 at 9:03 am
With a cluster, the fact you have 3 nodes is largely irrelevant.
The nodes form part of one cluster. All references should be made to the virtual cluster server itself.
Thus, all servers will be effectively subscribers to themselves. As far as I can see, there's very little (if any) benefit to having the style of setup you seem to be proposing.
Clustered SQL will happily switch over replication to the quorum holder of the SQL replication resources in the event of failure of the owning server (that's how I have it configured here)
The best place to replicate data to is a server outside the cluster, and have the cluster for resilience purposes of the source data.
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply