Win2K3 - Active-Active Configuration Doubts

Hello,

I have some doubts about Active-Active and Active-Passive cluster config for
Win2K3 servers. We want to install Active-Active cluster in Win2K3-R2 env.

I suppose that there is not difference in installing/configuring Windows
Server 2003 for Active-Active or Active-Passive cluster. Its all how we
install and configure the application to work with these clustered servers.

If this is true, are there any good links or best practices from microsoft
or others to be considered for installing Active-Active cluster ?

Please guide.

Thanks in advance.

abc.

Active just means that a node is configured for a clustered resource. So
you could easily have multiple file share resources with separate network
names and disk resources on each node respectively. The trick is to ensure
that that application and performance load can be supported on a single
node in the event of failover. One application which would not be suited
for A/A is Exchange. But in general, should work fine.

Planning Your Server Cluster
http://technet.microsoft.com/en-us/l.../cc739132.aspx

Guide to Creating and Configuring a Server Cluster Under Windows Server
2003 (download)
http://www.microsoft.com/downloads/d...ed7-9634-4300-
9159-89638f4b4ef7&DisplayLang=en

Dave Guenthner [MSFT]
This posting is provided "AS IS" with no warranties, and confers no rights.

Hello Dave,

Thanks for reply and the links.

So it means, there is no difference in configuring Win2K3 cluster server,
whether for A/P or A/A.

We have 2 servers and want to install SAP (DB-SQL2005 and CI) servers as
clustered servers.

Thanks again.

abc.

"Dave Guenthner [MSFT]" wrote:

> Active just means that a node is configured for a clustered resource. So
> you could easily have multiple file share resources with separate network
> names and disk resources on each node respectively. The trick is to ensure
> that that application and performance load can be supported on a single
> node in the event of failover. One application which would not be suited
> for A/A is Exchange. But in general, should work fine.
>
> Planning Your Server Cluster
> http://technet.microsoft.com/en-us/l.../cc739132.aspx
>
> Guide to Creating and Configuring a Server Cluster Under Windows Server
> 2003 (download)
> http://www.microsoft.com/downloads/d...ed7-9634-4300-
> 9159-89638f4b4ef7&DisplayLang=en
>
> Dave Guenthner [MSFT]
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>
>

Correct. I would create a minimum of three resource groups in the cluster.

Default Cluster Group with it own Virtual Network Name, IP Address, Quorum
Physical Disk Resource, and MSDTC resource. Make MSDTC dependent on network
name and quorum disk.

Create a separate group for each SQL/SAP instance, each with their own
respective virtual network names and disk resources so that they can fail
over independently. Once complete, simply move the desired SQL group to
cluster node.

Make sure that each node could if necessary own both resources at the same
time from performance perspective.

Dave Guenthner [MSFT]
This posting is provided "AS IS" with no warranties, and confers no rights.

Hello Dave,

Thanks for reply.

I will create different resource groups and test it. I am facing a different
problem in allocating pagefile to SAN partition.

I have a SAN partition for SAP DB which is 300GB in size. i want to allocate
50GB of pagefile (swap space) to this disk. Is this possible ?

I have tried but after setting the pagefile in both the clustered servers on
this partition, it does not allocate any pagefile on it. The size of the
partition remains same (300GB).

Can you please guide on this ?

Thanks again.

abc.

"Dave Guenthner [MSFT]" wrote:

> Correct. I would create a minimum of three resource groups in the cluster.
>
> Default Cluster Group with it own Virtual Network Name, IP Address, Quorum
> Physical Disk Resource, and MSDTC resource. Make MSDTC dependent on network
> name and quorum disk.
>
> Create a separate group for each SQL/SAP instance, each with their own
> respective virtual network names and disk resources so that they can fail
> over independently. Once complete, simply move the desired SQL group to
> cluster node.
>
> Make sure that each node could if necessary own both resources at the same
> time from performance perspective.
>
> Dave Guenthner [MSFT]
> This posting is provided "AS IS" with no warranties, and confers no rights.
>
>

Hi ,

A pagefile can only be set on "fixed" disks. not on shared storage.

So if you point a san disk directly on your server you can put a pagefile on
it, but if you assign this disk as shared storage than the page file will
not apply on this disk.

or if you use a net share * disk the pagefile will not set on this disk.
The reason for this is easy lets say you are running node1 with the pagefile
disk ( pagefile = "memory extender" machine depended) after a fail over the
app and the pagefile would failover to node 2
node 2 will now handle the memory from node 1 , this will never work ! even
if you would the problems with a fail over will get bigger and bigger.

So IMHO the best way is give the cluster nodes a fixed san target and use
this as swap disk. Get more real memory and minimize the pagefile usage is
my opinion.
the blog askperf is a great place to start.

http://blogs.technet.com/askperf/arc...or-anyway.aspx

http://www.brianmadden.com/blogs/bri...ally-mean.aspx

Greeting,

Robert Smit
http://fiberman.spaces.live.com/




"abckid" <> wrote in message
news:AB4AC4C6-F6B8-4A34-9BF9-...
> Hello Dave,
>
> Thanks for reply.
>
> I will create different resource groups and test it. I am facing a
> different
> problem in allocating pagefile to SAN partition.
>
> I have a SAN partition for SAP DB which is 300GB in size. i want to
> allocate
> 50GB of pagefile (swap space) to this disk. Is this possible ?
>
> I have tried but after setting the pagefile in both the clustered servers
> on
> this partition, it does not allocate any pagefile on it. The size of the
> partition remains same (300GB).
>
> Can you please guide on this ?
>
> Thanks again.
>
> abc.
>
> "Dave Guenthner [MSFT]" wrote:
>
>> Correct. I would create a minimum of three resource groups in the
>> cluster.
>>
>> Default Cluster Group with it own Virtual Network Name, IP Address,
>> Quorum
>> Physical Disk Resource, and MSDTC resource. Make MSDTC dependent on
>> network
>> name and quorum disk.
>>
>> Create a separate group for each SQL/SAP instance, each with their own
>> respective virtual network names and disk resources so that they can fail
>> over independently. Once complete, simply move the desired SQL group to
>> cluster node.
>>
>> Make sure that each node could if necessary own both resources at the
>> same
>> time from performance perspective.
>>
>> Dave Guenthner [MSFT]
>> This posting is provided "AS IS" with no warranties, and confers no
>> rights.
>>
>>

Hello Robert Smit,

Thanks for reply. I have configured the SAN just as you have mentioned. I
wanted to know if it was possible to use shared drive or not, which is clear
now. Thanks.

Also the links are very informative.

Thanks again.

abc.

"Robert Smit" wrote:

> Hi ,
>
> A pagefile can only be set on "fixed" disks. not on shared storage.
>
> So if you point a san disk directly on your server you can put a pagefile on
> it, but if you assign this disk as shared storage than the page file will
> not apply on this disk.
>
> or if you use a net share * disk the pagefile will not set on this disk.
> The reason for this is easy lets say you are running node1 with the pagefile
> disk ( pagefile = "memory extender" machine depended) after a fail over the
> app and the pagefile would failover to node 2
> node 2 will now handle the memory from node 1 , this will never work ! even
> if you would the problems with a fail over will get bigger and bigger.
>
> So IMHO the best way is give the cluster nodes a fixed san target and use
> this as swap disk. Get more real memory and minimize the pagefile usage is
> my opinion.
> the blog askperf is a great place to start.
>
> http://blogs.technet.com/askperf/arc...or-anyway.aspx
>
> http://www.brianmadden.com/blogs/bri...ally-mean.aspx
>
> Greeting,
>
> Robert Smit
> http://fiberman.spaces.live.com/
>
>
>
>
> "abckid" <> wrote in message
> news:AB4AC4C6-F6B8-4A34-9BF9-...
> > Hello Dave,
> >
> > Thanks for reply.
> >
> > I will create different resource groups and test it. I am facing a
> > different
> > problem in allocating pagefile to SAN partition.
> >
> > I have a SAN partition for SAP DB which is 300GB in size. i want to
> > allocate
> > 50GB of pagefile (swap space) to this disk. Is this possible ?
> >
> > I have tried but after setting the pagefile in both the clustered servers
> > on
> > this partition, it does not allocate any pagefile on it. The size of the
> > partition remains same (300GB).
> >
> > Can you please guide on this ?
> >
> > Thanks again.
> >
> > abc.
> >
> > "Dave Guenthner [MSFT]" wrote:
> >
> >> Correct. I would create a minimum of three resource groups in the
> >> cluster.
> >>
> >> Default Cluster Group with it own Virtual Network Name, IP Address,
> >> Quorum
> >> Physical Disk Resource, and MSDTC resource. Make MSDTC dependent on
> >> network
> >> name and quorum disk.
> >>
> >> Create a separate group for each SQL/SAP instance, each with their own
> >> respective virtual network names and disk resources so that they can fail
> >> over independently. Once complete, simply move the desired SQL group to
> >> cluster node.
> >>
> >> Make sure that each node could if necessary own both resources at the
> >> same
> >> time from performance perspective.
> >>
> >> Dave Guenthner [MSFT]
> >> This posting is provided "AS IS" with no warranties, and confers no
> >> rights.
> >>
> >>
>