SAP Portal : Do we need the ERS Instance in a SGeSAP Package ?

[Raoul Shiro]

Hello,

I'm a SAP Basis Administrator, and we are about to install a SAP Entreprise Portal 7.0 on a HP Cluster using SGeSAP capabilities.

There are some points described in the HP documentation that still need to be clarified.
I hope I'll get some help here.
http://docs.hp.com/en/T2803-90013/T2803-90013.pdf

DB Instance and JCI (central services SCS) will be included in their own legacy package as it is necessary to achieve High Availabiltu in SAP Java stacks.

But regarding the ERS instance (replication instance) we have some doubts. Do we have to include it in a package ?

SAP recommanded us NOT to define a switchover group for the Enqueue Replication Server, which make sense, as it it not a single point of failure.

According to HP documentation there are two ways to configure the ERS (page 17 Designing SGeSAP Cluster Scenarios)

"SGeSAP offers the legacy package types rep and arep to implement enqueue replication packages for JAVA and ABAP.
SAP offers two possibilities to configure Enqueue Replication Servers:
1. SAP self-controlled using High Availability polling
2. Completely High Availability failover solution controlled"



We have decided to use the first possibility :
HA polling. In this case, do we have to include ERS in a package ?

Thank you very much for your attention and your help.
Best regards.
raoul

[Carsten Krege · Accepted solution]

My understanding is that the SAP self-controlled mechanism is a pure SAP solution that requires you to build and test various scripts so that SAP would know when the enqueue server (enq) is starting on the node where the replicated enqueue (rep) is running. Then SAP would stop rep. My understanding is that you would not need to create a separate rep package in SG with this setup.

SGeSAP's mechanism is smarter: You have an enq and a rep package. These packages have an EMS resource configured which ensures that enq and rep always start on different nodes. If enq does a failover, it would always start on the node where rep is running. The rep package would automatically stop and restart again on another node so that a new failure of enq would cause it to start on the node where rep is running now.

Not sure why SAP does not recommend to have your own rep package. It is not required to make it highly available, but you have to ensure that rep is always started on a node where enq does not run. The rep package helps you to have this dynamic setup.

Also the SGeSAP solution does not require you to write new scripts and care about the enque/enrep/hafunc_* parameters. It is a ready-to-use solution and the intelligence is in the SGeSAP provided EMS monitor cmenqord. Honestly I don't see a reason to use the SAP polling mechanism when you bought SGeSAP anyway already.

Carsten

[Raoul Shiro]

Thank you Carsten for your valuable answer.

To be honest, I didn't really understand the EMS monitor concept in the first place.
I thought that when using SGeSAP failover solution, we still had to write and maintain our own scripts in order to make ERS and SCS communicate ...

So obviously, SAP advised us to rather use their polling solution.

So thank you again for your clarifications. We will now carefully check the SGeSAP solution.