February 4, 2009 at 1:22 pm
for your 4 instances I'd probably set each to initially have max memory 3GB and min memory 2GB.
Note that should both your active nodes fail your passive will not be able to handle all the instances unless you upgrade the passive to 32GB ram.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
February 4, 2009 at 1:25 pm
2 - 3 % page file use is fine. SQL server will use allocated memory, 2.3gb free memory is fine - no worries there. free virtual memory relates to your page file - have you set it to a fixed size of say 4GB? free memory and free virtual memory are not the same.
[font="Comic Sans MS"]The GrumpyOldDBA[/font]
www.grumpyolddba.co.uk
http://sqlblogcasts.com/blogs/grumpyolddba/
February 4, 2009 at 1:55 pm
Thanks Collin,
We have the page file size as 2046MB, we did not increased its size so far.On node 1,we set the max memory as 12 GB and left 4GB for OS and now the sql instance using whole 12GB. How we would know if SQL instance need morethan 12GB? and also frequently getting High disk queue lengh on C drive alaram in monitoring tool. If we really have the problem with high disk queue length where can I get some kind of error message in sql server side like error log.
To set the page file size I followed the method2 in the link http://support.microsoft.com/kb/889654
and its coming around 7GB. So we are planning to increase the page file size to 7GB.
If we set the page file size to 7GB, I guess the error Free virtual memory is less will be resolved?
Thank You
Madhu
February 4, 2009 at 9:55 pm
Madhu,
I think the points of some of the others posters aren't being heard.
By setting instance 1 on node 1 to use 12GB of RAM, you cannot expect a successful failover of the instance to another server in the cluster.
All of the SQL Server instances need to have their max physical memory set to that any 1 server in your 3-node cluster will be running all 4 of your SQL instances.
Therefore, in a cluster where all three servers have 16GB, you would need to do something like this:
OS/Other Apps: 4GB
Instance 1: 4GB
Instance 2: 3GB
Instance 3: 3GB
Instance 4: 2GB
Total: 16GB
The way you are potentially configuring it is like this:
OS/Other Apps: 4GB
Instance 1: 12GB
Instance 2: 3GB
Instance 3: 3GB
Instance 4: 2GB
Total: 24GB ---> more than what you have.
This scenario would happen if 2 nodes of your cluster were to go down, and 1 node was expected to take the load of all 4 SQL instances.
Hope that helps clear up some of the confusion.
In the end, you really should consider getting additional memory for each of your systems.
Regards,
Steve
February 5, 2009 at 12:05 am
hi,
Therefore, in a cluster where all three servers have 16GB, you would need to do something like this:
OS/Other Apps: 4GB
Instance 1: 4GB
Instance 2: 3GB
Instance 3: 3GB
Instance 4: 2GB
Total: 16GB
The way you are potentially configuring it is like this:
OS/Other Apps: 4GB
Instance 1: 12GB
Instance 2: 3GB
Instance 3: 3GB
Instance 4: 2GB
Total: 24GB ---> more than what you have.
Here , I need to clarify.We have 3node a/a/p cluster setup.
On node1: 16GB RAM
Instance:Ins1...Max memory 12GB and left 4GB for OS
Node2: Passive with 16GB RAM
Node3: 16GB RAM
Instances: Ins1
Ins2
Ins3
Ins4
I need to set Max memory for above 4 instances on Node3.
Viewing 5 posts - 16 through 19 (of 19 total)
You must be logged in to reply to this topic. Login to reply