Load balancing doesn't work properly

[former_member209959]

Hi,

We have an strange issue in one system with the load balancing. There are six application servers with the same operating system, RAM, CPU, network, etc, but in one of them the workload in ST03 is very higher than the others, and most of the users are logged in it

For example, for dialog steps:

And for all kind of task:

Any ideas where We have to investigate to discovered the reason?

Regards

[Former Member]

Hello,

No problem with ST03, it is about Load balancing

"Logon load balancing increases the efficiency of various workgroups that have been defined, in terms of performance and system resource consumption, by distributing users across available application servers based on requirements for workgroup service and load sensitivity."

What kind of log on group is configured?
Did you checked SMLG ?

List of active SAP server instances and available LOGON groups

  lgtst -H <host name message server> -S sapms<SID>

Please refer to:

SAP notes:
51789 - Poor user distribution in logon distribution
26317 - Set up for LOGON group for autom. load balancing
1847056 - 'Load balance not working'

Wiki:

http://wiki.scn.sap.com/wiki/display/NWTech/Usage+of+thresholds+in+SMLG

You can check the logon load by displaying a list of the servers that are available and an overview of load distribution in your system.

1. Choose CCMS -> Configuration -> Logon groups -> Goto -> Load distribution.

This list also shows the current performance status of the application servers that are both assigned to logon groups and currently running. Every five minutes, or after every five logons, each application server writes its own performance statistics data to a memory-resident table on the message server. The current logon server is refreshed for each group.

2. To refresh the performance status of an application server, double-click an application server line in the Instance column.
   
   
   

BR

Tamas

[former_member209959]

Hi Tamas,

We have six logon groups (apart of SPACE), but all of them have the same caracteristiques. I executed lgtst program and everything is Ok. I also checked these notes and documents, but I could see nothing. In smlg --> load distribution, the AS number 2 always has the most dialog steps, users, but not always the most quality or response time:

Regards.

[manumohandas82]

Hi Maximino ,

You can  also find some details using st03n - > Expert Mode -> Load History & Distribution -> Users Per Instance

You can go back on "previous period " to see the details of previous dates . Double clicking on any instance will give you the details on the users .

By classifying the users you may be able to come to a conclusion whether it is dialog/RFC user  .

And if it is a normal user try to classify them into regions  ( If your system is globally used )  . If any particular region is hitting one server then it is most probable that  their sap logon pad is not configured with the message server .

Thanks ,

Manu

[manumohandas82]

Hi ,

Also check RZ12 whether the parallel generators are configured for all the servers and whether the dialog usage percentage for RFC users are more or less same for all the servers

Also check for the RFC configured on other systems to this system . In the RFC connections  on other SAP systems   is the load balancing used ?

Thanks ,

Manu

[former_member209959]

Hi Manu,

All the servers are configured in RZ12 and all of them have the same resources.

Theoretically all systems use the load balancing for the RFC connections. But if I don't find anything, I'll check if there are some systems that connect directly against this AS.

Regards.

[former_member209959]

Hi, I've found that in the others five AS, the transaction at the top of steps is: <number range buffer>, but not in the AS number 2:

Does anybody know what this mean?

Regards.

[divyanshu_srivastava3]

Either a network issue or users are going directly to this application server, not by LB.

You need to put a network trace 1st and users running on this server.