SAP ECC on Cluster

[Former Member]

Hi,

We are trying to install SAP ECC and EP with HA on Linux Cluster .

Here is an overview of our servers.

Physical node u2013 PRDN1 & PRDN2

Virtual hosts u2013 PRDCI &PRDDB

Here is how we did the installation.

1. ./sapinst with hostname u2013 PRDCI ( For ASCS)

2. ./sapinst with hostname u2013 PRDDB ( FOR DB instance)

3. ./sapinst on Physical node without any hostname (For CI so it got deployed on PRDN1)

4. ./sapinst on Physical node without any hostname ( For DI it got deployed on PRDN2)

Once the system came up below is the status in SM51.

CI - Dialog Batch Update Upd2 Spool ICM

DI - Dialog Batch ICM

Now we have some doubts which we couldnu2019t find answer to, weu2019d be very thankful if someone could answer.

1. If Physical node 1 ( PRDN1) goes down along with CI and ASCS moves to Node 2 I believe DI will then act as central instance, but as of now it doesnu2019t have update process, spool and ICM, what should be done about those?

2. Users should be allowed access ( If required ) through ASCS IP? Although weu2019ve plans to add Dialogue instances on different servers in future.

Regards

Abhishek

[Former Member]

Hi all,

Physical node u2013 PRDN1 & PRDN2

Virtual hosts u2013 PRDCI &PRDDB

Here is how we did the installation.

1. ./sapinst with hostname u2013 PRDCI ( For ASCS)

2. ./sapinst with hostname u2013 PRDDB ( FOR DB instance)

3. ./sapinst on Physical node without any hostname (For CI so it got deployed on PRDN1)

4. ./sapinst on Physical node without any hostname ( For DI it got deployed on PRDN2)

We completed installation following the above approach and used below file systems.

/usr/sap/SID - local on each node

/oracle, /oracle/client - on each node.

/usr/sap/DVEBMGS05 and /usr/sap/D05 created locally during installation on each node.

/usr/sap/ASCS00 is clustered and fails over with PRDCI.

/sapmnt/<SID>

Present scenario is that DB fails over fine without any issues.

Cluster resource for PRDCI has scripts which unmounts on /sapmnt/, /usr/sap/ASCS00 one node and then tries to mount everything on to other node also stop ASCS on node 1 and start ASCS on node 2.

This part is not happening because if CI and DI are up they are using /sapmnt/<SID>, and when i try to manually failover PRDCI to node to then it unmounts /sapmnt/ from node 1, this leaves CI in non working mode.

And on second node since DI is running and is using SAPMNT cluster fails to start on that node and disables.

Please could anyone advise on this.

Regards,

Abhishek

[sunny_pahuja2]

Hi,

You should give access to user through ASCS IP only so that load balancing should work. If you have plan to add more dialog instance then simple add those instance to logon group in SMLG and when you add logon entries to GUI then use this logon group which will equally distributed your load to all application servers.

For first question, you need spool and update work process on dialog instance as well. As if CI goes down , your DI will work as CI.

Thanks

Sunny

[Former Member]

Hi,

I have few Doubts.

1. If we've configured enqueu and Message server in ASCS, do you still need enqueue replication server on both the nodes?

2. Shared/movable Directories for example /sapmnt/<SID>, /oracle/SID/112_64/ can these be mounted simultanously on both the nodes, or we should mount on one and then export as NFS to other?

3.Our /oracle/, and /oracle/clinet are local directories, /oracle/PEP/* i.e. Should all subseqeunt directores are movable between nodes.?

4. In a setup of ASCS00, DVEBMGS02 & D02 /usr/sap/<SID>/ASCS00 should be movable ,nothing else in the /usr/sap/ directory should be moveable, is that right?

Regards

Abhishek Sharma

[sunny_pahuja2]

Hi,

> 1. If we've configured enqueu and Message server in ASCS, do you still need enqueue replication server on both the nodes?

In Unix and Linux system, message server and enqueue both are part of the ASCS and it resides on disk that move across 2 nodes. So, there is no enqueue replication server separately.

But if you want to create enqueue server separately then you need to do it manually.

> 2. Shared/movable Directories for example /sapmnt/<SID>, /oracle/SID/112_64/ can these be mounted simultanously on both the nodes, or we should mount on one and then export as NFS to other?

why you want /oracle/SID/112_64/ to be mounted on both nodes ? Install it on movable disk. so that if CI node is not available then it will move to 2nd node.

/sapmnt/SID should not be NFS mounted. Because if you will NFS mount from node1 to node2 and in case node1 will go down then NFS mount will also not available on node 2. You should simultaneously mount it on both the nodes. You can discuss the same with your hardware team then will do it.

> 3.Our /oracle/, and /oracle/clinet are local directories, /oracle/PEP/* i.e. Should all subseqeunt directores are movable between nodes.?

Oracle software and database should be installed on movable file system. If you want to install oracle software locally then you have to install it on second node as well.

> 4. In a setup of ASCS00, DVEBMGS02 & D02 /usr/sap/<SID>/ASCS00 should be movable ,nothing else in the /usr/sap/ directory should be moveable, is that right?

>

Right.

Thanks

Sunny

[vincentlim826]

Hi,

1.you are running standalone enqueue with mesg service architecture, CI will be just a normal instance.Your main central service is at ASCS. You still need application server in node 2 to do update/spool service.

2. I believe so. ASCS contains enqueue/message service.

Regards,

Vincent