Replication server not starting after node-1 goes down

[Former Member]

Hi,

I have installed SAP R/3 4.7 on Oracle 10g RAC. Following are the details about the Environment:

OS: RHEL 5.3

Database: Oracle 10.2.0.4

FileSystem: OFCS2

SAP Version: SAP R/3 4.7 on 640_EX2 Kernel.

HA Monitoring Tool: SAPCTL 4.0

RAC Setup: 2-Node Cluster with Oracle clusterware(CRS)

Problem:

I checked both the Enqueue server and replication server logs and they are connected with each other. For testing this HA setup, I shutdown the node-1 where Enqueue server is running, the user VIP for abapenque is not restarting on node-2 where the replication server is running and waiting for enqueue server to start.

I checked the replication server logs:

[Thr 47225998538864] Tue Sep 22 10:49:47 2009

[Thr 47225998538864] *** ERROR => EncNiConnect: unable to connect (NIECONN_REFUSED) [encomi.c ; 445]

[Thr 47225998538864] *** ERROR => RepServer: main: connect failed with return code -7 [enrepserv.cp 587]

which is fine since the connection is lost.

The Replication server and Enqueue server are configured using SAPCTL 4.0 utility.

Questions:

1. The VIP should failover to node-2 and Enqueue server should start on node-2. Why the VIP is not failing over after replication server reported the lost of enqueue server connection?

2. The Replication server remains in the 'hold' state and waits for enqueue server to respond, any parameter that we can set to reduce this waiting time?

3. If any one can share the setup of start and instance profile for ASCS and ERS would be very help full.

Thanks,

Raj.

[brian_walker · Accepted solution]

Do you have something which detects that the enqueue server has stopped on node-1 and attempts to start it on node-2? As far as I know and have experienced, SAP does NOT provide this functionality themselves. The failover of the ASCS/SCS is completely up to the hardware/OS vendor for the platform you are on. In the Windows world, this is done with MSCS and when an HA system in installed in an MSCS, the SAP installer configures MSCS so that it will start the ASCS/SCS on another node in case of a failure.

For non-Windows (aka Unix/Linux), it is up to YOU to configure this. As you may have already noticed, for non-Windows SAPINST does not even have a step for installing the enqueue replication (ERS) and that must be done manually (There are some somewhat helpful ERS instance profile examples on help.sap.com). The same is true for the failover setup. On some platforms (IBM zSeries for instance), this is accomplished with Tivoli System Automation.

If you just wanted to test that the failover would work (without automation of course), you can make sure the enqueue replication is running on node-2, stop the ASCS/SCS instances on node-1, and start the ASCS/SCS instances on node-2. What should happen is that the enqueue table is rebuilt from the replication server and then the enqueue replication instance on node-2 will stop. You will then need to start the enqueue replication instance back on node-1 so that another failover test from node-2 back to node-1 will work correctly.

Brian

[Former Member · Accepted solution]

Thank you so much for the reply,

Update:

Yes, We have Oracle Clusterware to Monitor the start,stop and check actions on Enqueue and replication server.

The SAPCTL utility is used to start, stop and relocate the abap enqueue and replication server between the nodes during the failover and switchover.

Suggested Inputs for help.sap.com:

I was reading about HA Pooling concepts, but can not find the script which I have to put in the parameter ( Which should be given by HA Vendor, i.e. Oracle Clusterware in my case):

enqueue/enrep/hafunc_implementation =script

Question:

1.Is any one made the start,stop and check scripts for enqueue and replication server in Oracle RAC Environment?

Or

2.any Work around to notify the CRS to start Enqueue server on node-2 after we shutdown the node-1?

Thanks,

Raj.

[Former Member · Accepted solution]

Hi,

did you use the SAPCTL white paper: http://www.sdn.sap.com/irj/sdn/go/portal/prtroot/docs/library/uuid/d2827bef-0c01-0010-75a3-f96aa47fe...

If yes, you don't need to use Self-Control with HA poling and thus no enque/enrep/hafun since this is covered by SAPCTL based on Oracle clusterware and is responsible for failing over the VIP and services

You have to use the start and instance profile for ERS based on SAPCTL documentation

regards

Barthez