MSCS questions

[Former Member]

Hi all,

I installed a MSCS ABAP ERP 6.0 in Wins 2k3 + SQL Server 2k5 with 2 nodes (Node1: ASCS, DB instance, CI; Node2: ASCS, DB instance, Dialog instance).

But I have some questions below:

1. About the 3 hostnames for 3 virtual components mentioned in the installation guide: "clusgrp" for component "Cluster group", "sapgrp" for component "SAP cluster group", "dbgrp" for component "DB cluster group". I think the name "... group" makes confusion, at the beginning it made me feel they're the cluter group names in MSCS. In fact, the "Cluster group" is the MSCS cluster name when installing MSCS. The "DB cluster group" is the virtual server name of SQL Server. The "SAP cluster group" is the Virtual Instance Host of SAP. Anything is wrong?

2. When installing CI & Dialog instance, the installation guide says that we have to install them on local disk, but I find no screen which tells me about this. I don't know why.

3. About the instance numbers, I chose the default number that SAPinst raised. That's is: 00 for ASCS, 10 for Enqueue Server Replication, 01 for CI, and 01 also for Dialog instance. My SAPGUI system uses 01 as the System Number and for me it's ok with the failover. Anything is wrong?

4. I want to install another system for Dialog instance. This system will be the node out of MSCS. What does it need? 1 local disk for OS only? 1 public net card only? How about its instance number - I have to use another number except 00, 01, 10 such as 02?

Thanks,

Toan Do

[Former Member · Accepted solution]

Toan,

Your answers below:

1) This make sense if your Db and SAP Application are on different servers - review the distributed installation and it will make sense. It is typical for larger installations not to co-locate the CI and Db applications on the same server.

2) Yes - local disk - installation script assumes that you read and complied with the installlation guide which prepares the installation - the install script does the installation. Why?? If you have a network issue and try to write or read from the installed file system and it fails, the system fails so local disk only. Dialog instances actually have some cool things they do to keep this situation from occuring even though they use shared file systems somewhat.

3) Instance numbers only help to separate the ports that systems on the same box need to use - just don't re-use the same system number on the same host (installation script should keep you from messing this up).

4) Read the insallation guide on installing a dialog instance - you have to share some file systems for this to work right and you have a number of steps to complete or you will have some real complex problems.

Hope that answers your questions, please set this message to answersed if it has.

[Former Member]

Hi Tomas & Raj,

Thanks for your info., but fortunately, my ERP 6.0 SR2 in Wins 2k3 Sp2 + SQL Server 2k5 SP2 doesn't face this case. My saploc points to local disk.

Toan Do

[Former Member]

Good, because the problem I described is not funny at all and at that time you have to throw away the installation guide since this is not described anywhere.

Let's go on with your questions:

1)

You are right

2)

since SAPLOC points to the local disk on both nodes no questions will ever be asked about location of the installation.

Still, look out for the SAPLOC in the SAP Cluster Group.

3)

This is correct and just a matter of taste (what instance number you choose).

4)

You need a local disk for OS, SAP executables and swap.

Depending on the hardware (32 or 64-bit) you'll different size of the swap file.

In 64-bit environment you'll need a minimum of 20 GB swap !

Also space for the SAP executables (have a look at the MSCS-nodes local \usr\sap directory and you get the picture).

If you also add the Java-stack, an 72 GB disk tend to be too small.

About the instance number.

It doesn't really matter what you choose as long as the instance number is unique on the host (hosts outside MSCS).

To make things consistent, you can choose the same instance number as for the DI on the 2nd node (01).

SAPGUI and http-access to any of the servers will then use the same port number.

[former_member229014]

This Bug is explained in section 5 of the Main windows Netweaver 7.0 instalaltion Note(967123). Please read the installaltion notes for windows and SQL serer before starting the installation.

You must install the DI and CI instalaltion in the local drive only as specified in the installation guide.

[Former Member]

Hi,

0. About 2 nodes (Node1: ASCS, DB instance, CI; Node2: ASCS, DB instance, Dialog instance) => That's from the the figure of Installation Guide where ASCS & DB are failed over bet. node1 & node2 and CI is installed on node1 while DI is installed on node2. I didn't mean that I installed ASCS & DB instance on node1 and also on node2, I followed the installation steps.

1. About the 3 hostnames for 3 virtual components mentioned in the installation guide => I think I'm right.

2. About when installing CI & Dialog instance, the installation guide says that we have to install them on local disk, but I find no screen which tells me about this => But I find out a folder /usr/sap was installed on local disk (C:).

3. 4. About the instance numbers => I find out it's ok if not choosing the same instance number which declared before on the same host.

[Former Member]

Hi,

Regarding your Question 2:

SAPinst first looks for a share named SAPLOC.

If it exists, SAPinst will install SAP at that location.

If it does not exist, SAPinst will ask you for the location.

I have seen som strange behaviour from SAPinst in an MSCS-environment.

I will try to explain the problem in my case:

!! Note that this has only been observed in MS-SQL environment !!

SAPinst places (by mistake) SAPLOC in the same group as the SAP Virtual Hostname (together with SAPMNT).

this makes the CI to be installed on the shared disk instead of the local disk on the 1st node.

Then, the installation of the DI is installed locally on the 2nd node

(since there was no SAPLOC on the 2nd node at the time).

What happends when you then try to "roll" the Clustered-services (SCS) to the 2nd node.

I mean, you must check that the fail-over really works.

What happens is the following:

Not CI nor DI will start again if you stop them.

Why ?

Because both SAPLOC and the shared disk missing from the 1st node.

On the 2nd node SAPLOC points to the shared disk, which is wrong for the 2nd node.

Now, we roll back the SCS to the 1st node.

What happens is the following:

CI will start but not DI.

Why ?

Because both SAPLOC and the shared disk is back on the 1st node.

On the 2nd node SAPLOC does not exist so the DI can not start.

In the end we got ourselves and MSCS-installation which can only run on the 1st node.

the 2nd node is useless.

In order to prevent this you have to check the SAP clustergroup for the share SAPLOC before

starting the CI or the DI installation.

If it does exist, remove it.