June 16, 2008 at 3:36 am
I am currently involved with a production environment that is made up of a two node Acitive/Passive cluster.
I am conscious (and so too of course is the business) that the Passive node is doing absolutely nothing, all of the time other than of course waiting to take over in the event of a failure.
My question is would it be possible/wise to install a secondary named instance on the Passive node to be used for non-essential processing, such Reporting and Business Intelligence, thereby making use of the available processing power not currently being utilised.
Also, we know that a single node could comfortably support both instances of SQL Server in the event of a failover but I would have no issue with immediately stopping the named instance in order to grant full resource to the production instance.
I would really be interested to know your thoughts on using a configuration of this nature and perhaps some of you are even currently working with similar environments?
Many Thanks,
John
June 16, 2008 at 6:08 am
The answer ..... it depends
You can install an instance on the passive cluster, but as long as it's passive, it can't see the data from the active cluster. So you'll have to store these databases locally, or on a separate SAN/NAS/... storage.
If the passive server becomes active, it will share memory and CPU for both instances (assuming you did not install procesor affinity). Thus, make sure you'll have enough resources for your failover environment.
What most people don't realize (i was one of them until somebody told me this) is that it's possible to use both servers in your environment and make each server a failover for the other. Server A has application X, with Server B as a failover. Server B has application Y with Server A as a failover.
This is quite easy to implement this, with some additional design about failovergroups etc.
In this case you'll use both machines in your environment. If you have a failover, both applications will run, but at a slower speed.
But ask yourself: how often will a failover happen compaed to the use of a failover in this scenario?
Wilfred
The best things in life are the simple things
June 16, 2008 at 6:21 am
Hi,
Thank you for your reply.
It sounds as if we are thinking along the same lines. I was planning on implementing a separate SAN if we were to go ahead and install an additional SQL Server instance on the Passive Node.
An Active/Active configuration would not be necessary in this case as the applications (i.e. Reporting Services/Analysis Services/ BI etc.) to be served from the Passive Node are not essential services and so can be unavailable for quite some time without issue.
Many Thanks,
John
June 16, 2008 at 7:03 am
The short answer is yes, you can use the passive node. The rule of thumb is to make sure there is enough overhead on all physical nodes such that if all resources were failed over to each one individually, you aren't in a memory, CPU, or I/O short situation. Even though you indicate it'll be used for non-essential, there may be times where it will fail-over and you won't catch it and people are still using the "non-essential side," so you want to make sure you're fine resource wise.
K. Brian Kelley
@kbriankelley
June 16, 2008 at 7:16 am
Thaks for your comments.
June 16, 2008 at 11:54 am
yes it is defenitly possible. And it is always a good idea to run the reporting instance in the passive node as it will solve the n/w issues for replication (being both the nodes in the same LAN).
Only thing you need to take care is you will have to have a seperate Virtual Name, Virtual IP for the new cluster. Also make sure that each node got enough capacity to run both the instances (in case one of them fail over to another).
As SQL Server 2005 manages memeory dynamically (even if AWE enabled for higher memory) you need not specify the max and min memory options.
subban
June 17, 2008 at 2:02 am
Hi Subhash,
Thank you for your comments.
I am not sure if we would need to create a new Virtual Name and IP. I was under the impression that we could simply add another named instance onto the Passive Node and still reference it using the current local server IP and the instance name?
Good points regarding the memory. I am thinking that I will limit the resources available to the named instance in order to ensure that there will always be adequate resource available in the event of a failover from the Active node.
Many Thanks,
John
June 17, 2008 at 6:40 am
You may also want to consider that you don't have to license a passive node. If you put an instance on the passive node, you'll have to license it.
Tim White
June 17, 2008 at 6:44 am
Good point, although given the amount of licenses we have, one more will make little difference :-).
June 17, 2008 at 6:54 am
BigJohn (6/17/2008)
Hi Subhash,Thank you for your comments.
I am not sure if we would need to create a new Virtual Name and IP. I was under the impression that we could simply add another named instance onto the Passive Node and still reference it using the current local server IP and the instance name?
If it's a named instance (data engine), you will need a new virtual name and IP.
K. Brian Kelley
@kbriankelley
June 17, 2008 at 6:56 am
Thanks Brian.
June 17, 2008 at 8:37 am
Back to the licensing issues for a second. Are you using Enterprise or Standard Edition on the cluster? You may not think licensing is much of an issue incrementally, but Enterprise is not cheap.
We're running 2 clusters: 1 prod 4 node/ 3 instance cluster and 1 dev 2 node/2 instance cluster. I saw a comment about setting up a separate SAN if you're going to add a second instance to your cluster. This is not necessary. You'll simply need to carve out more space as logical volumes to assign to the SQL resource group(s) for the new instance on your current cluster. We have 2 SAN environments at separate physical locations. All nodes of prod run off 1, all nodes of dev off of the other.
- Tim Ford, SQL Server MVPhttp://www.sqlcruise.comhttp://www.thesqlagentman.com http://www.linkedin.com/in/timothyford
June 17, 2008 at 9:03 am
Hi Tim,
Thank you for your comments. Standard edition is what the environment will be running.
We also have plenty of room for expansion on the existing SAN, so as you say a separate logical partition should do the trick.
Cheers,
John
Viewing 13 posts - 1 through 12 (of 12 total)
You must be logged in to reply to this topic. Login to reply