March 4, 2010 at 9:37 am
Hi there,
We are looking at building a 2008 2 node cluster for one of our applications. This would be our first SQL Server cluster, so we are trying to get up to speed on the topic.
We currently are running an Active/Active 2 node Oracle cluster for a different application, and it works great. One single node is enough to manage all of the work load, but by having both clusters active the work load is split between them by sort of round robining the connections. When a failover occurs there is almost 0 down time because the other server takes over so quickly it's practically seamless to the end users. The worst we've seen is maybe a dropped connection, but the end users are able to re-log in with out a problem.
Would the Multi Instance mode work the same way without having to down any services? From my limited reading so far I'm getting the impression that 2 nodes cannot both be a part of the cluster, and instead they are meant to automatically just take over the database in the event of the primary node failure.
I'm not certain if I'm clearing stating my question, but we're looking to build the 2 nodes in a way the obviously gives us redundancy as well as the least amount of outage in the event of a fail over.
March 5, 2010 at 2:04 am
The duration of the 'outage' on any Windows cluster is the sum of the time it takes to detect a failover is required plus the time it takes to unmount and remount the LUNs plus the time it takes to start the services on the failover node. In the case of SQL Server this is usually under two minutes..resuming operation via the same network name and IP.
There are other ways that might save a little time but maybe not enough to be worth the effort..like peers behind a load balancer...implies a tiny bit of latency between the peer data...or a mirror with a witness and a load balancer. Note that .NET connection strings can be configured with two sql instances...can attempt a connection to a mirror if the primary cannot be reached...just read about that a few days ago.
March 5, 2010 at 2:18 am
i would suggest you for single instance failover cluster rather than multi instance cluster.
for more details you can refer these
SQL Server 2005 Failover Clustering White Paper
How to: Create a New SQL Server Failover Cluster (Setup)
Regards,
Sarabpreet Singh 😎
Sarabpreet.com
SQLChamp.com
Twitter: @Sarab_SQLGeek
March 5, 2010 at 6:53 am
Thanks for the information. We'll probably end up doing the active / passive approach since there wasn't really a hard demand for us to provide the seamless failover. I just wanted to be sure.
Is there a SQL server technology that mimics the load balancing approach? Where we would have multiple nodes all sharing the workload on a single database? I seem to recall something many years ago called Federation or something like that.
Thanks again
March 5, 2010 at 10:35 am
tluker (3/5/2010)
Thanks for the information. We'll probably end up doing the active / passive approach since there wasn't really a hard demand for us to provide the seamless failover. I just wanted to be sure.Is there a SQL server technology that mimics the load balancing approach? Where we would have multiple nodes all sharing the workload on a single database? I seem to recall something many years ago called Federation or something like that.
Thanks again
No - there is no load balancing clustering technology available for SQL Server. And, I don't think it is something that is going to ever be considered for SQL Server.
Clustering for SQL Server is only used for HA scenarios.
Jeffrey Williams
“We are all faced with a series of great opportunities brilliantly disguised as impossible situations.”
― Charles R. Swindoll
How to post questions to get better answers faster
Managing Transaction Logs
March 5, 2010 at 1:42 pm
Wow, that just seems hard to believe that there wouldn't be any desire to scale SQL Server in that way. I have to think there is a lot of money in the ability to scale to multi node processing, and that microsoft would be looking to move towards a solution that will work the biggest of clients.
Isn't there some sort of new technology that is being introduced in R2 that has to do with scaling data warehouses?
March 5, 2010 at 2:07 pm
You can accomplish some of the same things but a lot depends on what you are doing. Data updates tend to need to be done at one place or given some other methods can be spread out but you have to plan, plan, plan the whole thing.
What is the problem you are trying to solve instead of coming at it from the solutions direction, lets here about your problem..
CEWII
March 5, 2010 at 2:15 pm
SQL Azure is the latest major investment in database scaling technology. There is also upgraded peer replication that adds geo-distributed synchronized sql database capability as well.
March 5, 2010 at 2:34 pm
Well, I think because we are only doing a failover cluster that doesn't have a requirement of 99.999% (or whatever) up time, we'll be good doing the active/passive plan.
I am new to SQL Server clustering, only having experience with our 2 node Oracle Active/active cluster. The thing that I like about that solution is that for instance we had a node go down last week, and nobody noticed an outage, because those on the node that went down failed over to the other active node so quick, that they didn't know there was a problem. It's possible someone noticed it, but we didn't get any reports of it. Had it been a case where the failing node had to shut down, and the failover node had to come up, then there would have been maybe a minute or two outage, then I can guarantee would have produced many system failure requests. Then we would have been having to respond to all of those call tickets, and directors wanting to know what the issue was, as well as just reacting to the failed node.
We were contimplating using the multi server (active/active) thinking perhaps we could get closer to having that minimized down time by having the secondary node's SQL Service to the failing node already running. We were thinking/hoping that by having both SQL services already running, it would reduce the amount of time for the failover node to take control over the database from the failed node. In the end reducing the amount of people that would notice an outage, thus reducing the amount of reports of an issue we need to follow up on.
Thanks again for any recommendations. This is turning into a very good learning experience, which I am enjoying.
March 5, 2010 at 2:41 pm
Thanks SSC Rookie, I believe that is what I heard about.
I knew Microsoft wouldn't let that get by them. Every article I ever see about Enterprise RDBMS's always refers to Oracle, and DB2, and SS seems to be considered a mid size company solution. From all the improvements I've seen to SQL Server through the years, it seems obvious they have every intention of being considered just as big of an enterprise player.
Go SS!!!
Thanks again.
Viewing 10 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply