ECC 6.0 on Windows Cluster

[Former Member]

Hi Everybody !

Regarding ECC 6.0 Installation on Windows MSCS, the ERP SR3 Guide says that :

- Central Instance and Dialog Instance are not Single Point of Failures and they are installed locally ...

So, the Q is what if the Central Instance Goes Down ? How do we assure that there is no downtime / minimum downtime ?

'Cause, the CI is not in Cluster Fail Over Scope , there is no 'Standby Option ' for this C I and Dialog Instance .

Can anybody pl. throw some ' Lights ' on this ? Appropriate Reward Points Guaranteed in addition to sincere appreciation of your

time and inputs.

Thanks a Lot in Advance,

Regards,

- Ishan

[Former Member · Accepted solution]

For ECC 6.0, the central instance is just a normal instance. It does not have an ENQ process.

The ENQ process and message servers are moved outside into the SCS (Central Services) instance. This instance is just two running processes, message server and enqueue. This instance will be running on the MSCS.

Why a central instance w/ no enqueue? Not sure, it's SAP requirement. What's really going to piss you off is that not only is this the case, what's even worse is the "SAP supported" way is to run this instance on local drive (instead of a shared drive on the MSCS).

So you end up w/ this:

SCS (message server + enqueue) running on shared drive, part of MSCS.

A central instance install on each MSCS nodes, w/ a kernel + work folder in each node's physical drives.

What did we do in-house to fix this? Well, we didn't follow SAP's suggestions. Instead, we created the central instance as a MSCS on a shared drive. This way, we have this:

SCS (message server + enqueue) running on shared drive, part of MSCS.

Central Instance running on shared drive, part of MSCS.

This way, we don't haev to support / monitor each local drive on each node.

[Former Member · Accepted solution]

Thanks a Lot, Kevin !

You are cent per cent right ...Keeping CI outside MSCS do not make any sense.

So, since you have already implmented the CI within MSCS ( which is not SAP Recc. Way ..)

- Does SAP Provides support for such implmentation or they have ' hands-off ' approach for this ?

- By any chance , do you have any links / blogs which would give details about how to implement

CI in the MSCS ? ..'Cause, it being ' not SAP Way ' ..obviously, I see only vague, brief mention

about CI implemenation in MSCS..

Thanks again,

Best Regards,

- Ishan

[Former Member · Accepted solution]

Not sure how much of the implementation I can talk about here, and I don't have a blog ... sorry.

Basically, to get a CI on the MSCS is as easy as pie.

First, the basics:

0. Create a SAP MSCS group.

1. Create a MSCS shared drive and put the CI bits there (the whole usr\sap directory). Put this in the SAP MSCS group.

2. You will need a VIP (IP) and virtual name (Network Name) for the CI. These two items should also be cluster resources. Put these two items in the SAP MSCS group.

3. Check the user environment variable on the user that'd be running your SAP NT services. Make sure SAPLOCALHOST and SAPLOCALHOSTFULL is set to the virtual name.

4. Check the start and instance profile, make sure that SAPLOCALHOST and SAPLOCALHOSTFULL is set to the virtual name.

Now, create the SAP Cluster Resource type:

1. You need to do this on every node in your MSCS.

2. Go to cluster admin and right click. See if you can create a cluster resource type "SAP Cluster Resource" or something that starts with "SAP". If you see it, you are ok, skip this step. If you don't see the SAP Cluster Resource type, on every kernel download, look for a folder called cluster. Inside of this subfolder are 4 files you need: insaprct.exe, SapClus.dll, Saprc.dll, Saprcex.dll. Copy these files to the c:\windows\system32 folder and run insaprct.exe. After this, you should see the SAP Cluster Resource type. If you don't, check to see the kernel matches the O/S (32bit vs 64bit, etc.).

Now create the SAP NT service

1. Failover the SAP MSCS group to node 1.

2. Run the sapstartsrv.exe on the shared drive to create the NT service (SAP<SID>_<SystemNumber>).

3. Manually create saposcol by runnning: sc
node1 create "SAPOSCOL" binPath= "shareddrive:\usr\sap\SID\...kernel\saposcol.exe service"

4. Change the SAPOSCOL NT service login credential to the same domain account running the SAPSID NT service. Set to manual start.

5. Change SAPSID NT service to manual start.

6. Go to cluster admin and right click. Create a generic NT service for the SAPOSCOL. Create a SAP Cluster Resource for the SAPSID NT service.

7. Failover to each of the additional MSCS nodes and repeat steps 2, 3, 4, 5. You don't need to repeat 6.

As for SAP's support on this, I don't know actually. We have actually not had to depend on SAP Basis support for a while. But it's easy enough, just NT services controlled by Cluster Admin. The more you do it the better you become