Central Instance failover on NW70 WIN2008 Cluster

[paulo_sousa3]

Hello Experts,

after reading a lot of forum threads about Netweaver 7.0 Central Instances in cluster, it's not yet clear for me how it works.

If there is only one CI in the system installed on a "local disk" of the cluster, how is it protected from failure? In 4.70 I had the CI on a shared disk with a virtual name. I could failover it from node to node with no problem. How does it works in NW70?

The Dialog Instance installed in the other cluster node can work as a CI if I shutdown the real CI on the first node?

If this is true, I need to change the profile of the dialog instances to have all the required processes needed by the system

to work properly (UPD, SPO...)

Let's have an example:

I finished to install my WIN2008 2 nodes cluster with SQL2008 DB instance and also installed the Central Services (ASCS) instance on a shared disk. And now? If I understood I should:

1) install the CI in one node of the cluster and a Dialog Instance on the other node.

2) change the DI profile to have the some number and type of precesses running on the Dialog and on the CI.

3) create a logon group with both the CI and the DI and any aditional Instance running on external servers.

4) configure the SAPLOGON to use the logon groups specified on SMLG, using as "Message Server" my ASCS virtual name.

Is this ok? This means that I will no more have a virtual server name available to my Sap server. Is this right?

Thank you

[Former Member · Accepted solution]

Hi Paulo,

I am with you. If you are thinking in the perispective of architecture until ECC 5.0 or NW 2004, I feel it is not unusual to raise doubts you have.

Pls read the info from following links. Should clear your doubts.

If you still have doubts, pls do not hesitate to ask further questions.

Regards,

Krishna

[Former Member · Accepted solution]

Further , Actual Answers to your questions.

Q) The Dialog Instance installed in the other cluster node can work as a CI if I shutdown the real CI on the first node?

If this is true, I need to change the profile of the dialog instances to have all the required processes needed by the system

to work properly (UPD, SPO...)

Ans: It makes easy to understand if you call call it as Primary application instance and additional application instance than CI and DI. Which means affectivtely, your additional application instance or DI can take over and perform services of Central instance ( Primary instance) if it fails as the shared disk (S drive?) moves from node A to node B.

Q) install the CI in one node of the cluster and a Dialog Instance on the other node.

A. Yes and ASCS on Shared drive.

Q) change the DI profile to have the some number and type of precesses running on the Dialog and on the CI.

A) Correct. No difference apart from instance names etc..

Q) create a logon group with both the CI and the DI and any aditional Instance running on external servers.

A) Again yes.

Q) configure the SAPLOGON to use the logon groups specified on SMLG, using as "Message Server" my ASCS virtual name.

A) 100% correct.

This means that I will no more have a virtual server name available to my Sap server. Is this right?

In total you will have two virtual servers.

Krishna

[paulo_sousa3 · Accepted solution]

Krishna, your answers were very helpful.

To have clear all the required steps to put a two nodes cluster working in NW70, here is my roadmap. Please let me know if something is not ok.

- Install the ASCS in a shared disk (active/passive cluster mode resourse). Protected from failure, this instance will run under a virtual name and will hold the "message server" and "enqueue serve" services.

- Install the clustered database instance (shared disks with db virtual name).

- Install the Enqueue replication server (on local drive) in all cluster nodes.

- Install a Primary Application Instance ("Central instance" option in sapinst.exe) on a local drive of the first cluster node.

- Install an Aditional Application Instance ("Dialog instance" option of sapinst.exe) on a local drive of the second cluster node.

- Adjust profile parameters of both instances to have all required processes to run alone, if necessary. UPD, VB,VB2, SPO will not be available by default in the DI installation.

- Adjust memory parameters on both instances and in SQL server in such a way that both (DB and Sap Instance) can have enough resources to run in any single node.

- Create a "logon group" (SMLG) including both instances (and any aditional Instances running on standalone servers outside the cluster).

- configure the SAPLOGON to use the logon groups specified on SMLG, using as "Message Server" the ASCS virtual name.

With this configuration I can "switchover" the DB or ASCS (or both) instances from node to node, without loosing service vailability. This is usefull for maintenance operations on the cluster nodes. No data will be lost. Only need wait for users logout to shutdown the instance.

In a situation of a node failure, DB and ASCS will failover to the other node. Users working on the "failed" node instance will loose their work, but will be able to logon again on the other node, as it is in the some logon group.

Is this ok?

thank you

Paulo Sousa

[Former Member]

Hi Paulo,

Dis you perform the steps you mentioned. I Have done the same steps and my DI/aditional application instance doesn't start when I'm testing the failover. The ASCS and SCS instances switch fine, but the DI does not start. Do I have to change the start profile of this additional application instance besides of updating the work processes?

Thanks and best regards,

Ana.

[Former Member]

Hi Paula,

I will suggest you to go through SAP guide with correct SR release you are going to install. This will tell you exactly where and on which node CI or DI has to be installed.

And installation of DI is option in case of cluster installation and yes you have to use VIRTUAL name with correct host file entrie to fail-over.

And before installing SAP make sure you have installed SQL properly on shared disk and check that resurces are failing over.

Regards

[paulo_sousa3]

the DI is optional, but if not installed I will have only one instance, and no redundance in my system.

I'm using virtual name to the DB instance, but in NW70 I have no way to create a virtual name to the sap instance, do I?

I have to achive redundance by having more than one sap instance running, grouped on a logon group. Am I right?

--Paulo sousa