Showing results for 
Search instead for 
Did you mean: 

AIX 5L's HACMP on Database Servers

Former Member
0 Kudos

Hello everyone,

I've been looking around for information on a 4 node SAP ERP/Oracle RAC architecture for AIX 5L and I'm being constantly lead to vendors being responsible for the clustering requirements.

I'm trying to setup a 4-server ERP architecture. Based on documentation provided in the SAP service marketplace, IBM's General Parallel Filesystem (GPFS) and PowerHA/HACMP is needed.

The following is our Production architecture...

2 Application Servers for SAP ERP Applications (clustered active-active)

- and -

2 Database Servers + External Storage mounted identically onto both (RAC Clustered)

I'm not sure how GPFS and PowerHA/HACMP will be distributed onto the 4 servers.

I would like to humbly ask the SAP implementation experts to please confirm this setup I've planned.

2 App Servers = PowerHA/HACMP

2 DB Servers = GPFS + Oracle Clusterware

Is it a requirement to implement PowerHA/HACMP and GPFS on all 4 of my servers?

Help would be greatly appreciated.

Many Thanks,


Accepted Solutions (0)

Answers (1)

Answers (1)

Former Member
0 Kudos

You don't need HACMP on Oracle RAC. CRS acts as HA for your DB nodes. If one node goes down CRS failover the services to the other node. The way we have done is :

ASCS (Enqueue and Message Server) two nodes using HACMP. One active and one passive with enqueue replication enabled.

Three SAP application servers with no HA. If issues arise with the node or the frame we use live partition mobility.

DB Server: Three RAC node with GPFS. CRS acts as a HA agent as well.

Its always a good idea to have a same setup in one of the other environments that you are planning to have for your PRD system. We have our QA system setup the same way as our PRD. This helps us in testing any upgrades (OS, Filesystem, CRS, Oracle, SAP etc.) on our QA systems first before introducing them to PRD. And even helped us during our initial setup for RAC and HA and also testing our failover capabilities interms of what if one of my RAC nodes goes down or if ASCS nodes goes down. Also having QA close to the configuration of what PRD has helps when doing performance testing.

Hope this helps.