Dear Expert
I'd like to listen your opinion about RFC load balancing.
We use SAP CRM 7.0 EHP2 include of NetWeaver 7.3.
Our server is consist of DB & Message Server and 3 AP Servers.
We have a many of RFC inteface clients to connect CRM server. RFC interface accounts are about 40~50s.
Three days ago, we have a problem in 3th AP Server server which shared pool was out of order so 3th AP server
can not provide service well. It worked very slow because of memory bottleneck.
When I check system performance and load balancing in SMLG, AP#1 and AP#2 Server's respone time was under 1,500ms and
AP#3 server's response time is reached about 30,000.
Fotunately most incomming RFC sessions are allocated to AP#1 and AP#2 by message server properly except 'ABC' interface.
At that time, 'ABC' RFC sessions were trying to connect to AP#1, AP#2 and AP#3 server both all.
Of course, a many of sessions or transaction were failed or processed slowly.
We use load balancing type is 'Best Quality' and it is checked option 'Ext. RFC-enabled'. Choosing time for searching Best Quality AP is running every 5 minutes. (rdisp/autoabaptime, 300)
First time, I doubted whether 'ABC' client doesn't use SMLG or will reuse sessions without disconnection.
But I could not find any evidence related above. 'ABC' client is running HP-UX and it uses SAP RFC library 7.0
======================================================================
> what librfccm.so | grep LIBRFC
LIBRFC (c) SAP AG: Version: 700 Patch level: 0 Patch number: 199
thread-safe build 64 bit
======================================================================
Because SAP RFC library was released in 2009 year, so I was thinking RFC has a bug.
I almost read SAP notes but I didn't find any related Note.
How can I understand this situation?
Best Regards