Synchronous messages taking more time (than usual) during processing in AE

[Former Member]

Hello Experts,

We have 1 time-critical real-time interface, for which the processing needs to be completed within 15 seconds. If the response is not received within 15 seconds, then the process times out on the source system.

Source <--> SOAP sender (sync) <--> SAP PI <--> JDBC receiver (sync) <--> Target

The interface usually completes the processing within 2-3 seconds. But there are instances where the the user has reported time-outs and when we check the audit logs, it clearly indicates a processing lag within SAP PI system itself.

Below is the audit log information for one of the transaction that timed-out (meaning which took more than 15 seconds end to end)

27.06.2013 10:37:41.571 Information SOAP: request message leaving the adapter (call)
27.06.2013 10:37:41.571 Information The application tries to send an XI message synchronously using connection SOAP_http://sap.com/xi/XI/System.
27.06.2013 10:37:41.574 Information Trying to put the message into the call queue.
27.06.2013 10:37:41.577 Information Message successfully put into the queue.
27.06.2013 10:37:41.579 Information The message was successfully retrieved from the call queue.
27.06.2013 10:37:41.579 Information The message status was set to DLNG.
27.06.2013 10:37:57.968 Information The application sent the message synchronously using connection SOAP_http://sap.com/xi/XI/System. Returning to application.
27.06.2013 10:37:57.969 Information SOAP: call completed
27.06.2013 10:37:57.970 Information SOAP: completed the processing
27.06.2013 10:37:57.971 Information The message was successfully transmitted to endpoint http://xxxx:port/sap/xi/engine?type=entry using connection SOAP_http://sap.com/xi/XI/System.
27.06.2013 10:37:57.971 Information The message status was set to DLVD

I am just confused on how to troubleshoot this issue. From the logs, it looks like messaging system is taking more time.

Is there any particular reason why the PI processing takes more time between those highlighted steps?

If you could point me to some of the areas where I need to look at to find out the root cause, that will be really appreciated.

[ambrish_mishra]

Hi Prajeet,

Is it a test system or production system?

First try to decouple where was the time lag:

Was it data issue (data too big) ?

Was it in PI or in Database system. Look in RWB end to end monitoring, if it is not the database.

Check with PI admin if the system had hit peak volume/processing during the message processed.

If all of the above is negative, then you might have to look at system tuning and performance.

Ambrish

[ambrish_mishra]

Hi Prajeet,

Found a good document which should help you. Look for parameters specific to synchronous message processing.

http://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/f0b96675-87cf-2c10-b489-dee0ac03f...

Ambrish

[baskar_gopalakrishnan2]

>Is there any particular reason why the PI processing takes more time between those highlighted steps?

This could be due to various reasons.

a) Please check whether you see any slowness in the network connection between PI and target system. Check with network admins

b) Even if your data volume is more or any delay in data processing in the db side, you might find delay response from target system. Also check with target system dbas any reason for delay.  If you could able to access db system outside pi like TOAD or similar using tool, try to simulate the same data using sql query and see the response time. 

This is certainly not due to AE processing.

[Former Member]

Thanks for the reply, Baskar.

Even I thought that this was probably due to the data that was being requested from target through the interface (as it was random) or due to the network slowness.

But recently I found more specific logs that indicate otherwise. Let me share that with the group

Logs when the request is processed successfully within the time-frame

02.07.2013 10:44:58.586 Information The application tries to send an XI message synchronously using connection SOAP_http://sap.com/xi/XI/System.

02.07.2013 10:44:58.602 Information Trying to put the message into the call queue.

02.07.2013 10:44:58.605 Information Message successfully put into the queue.

02.07.2013 10:44:58.612 Information The message was successfully retrieved from the call queue.

02.07.2013 10:44:58.613 Information The message status was set to DLNG.

02.07.2013 10:44:58.614 Information Executing request mapping "Mapping name"

02.07.2013 10:44:58.648 Information Delivering to channel: In_JDBC

02.07.2013 10:44:58.651 Information MP: entering1

02.07.2013 10:44:58.655 Information MP: processing local module localejbs/CallSapAdapter

02.07.2013 10:44:58.656 Information JDBC Adapter Receiver processing started, required QoS BestEffort

Logs when the request takes more than 15 seconds

02.07.2013 10:45:43.690 Information The application tries to send an XI message synchronously using connection SOAP_http://sap.com/xi/XI/System.

02.07.2013 10:45:43.709 Information Trying to put the message into the call queue.

02.07.2013 10:45:43.716 Information Message successfully put into the queue.

02.07.2013 10:45:43.723 Information The message was successfully retrieved from the call queue.

02.07.2013 10:45:43.724 Information The message status was set to DLNG.

02.07.2013 10:45:43.725 Information Executing request mapping "Mapping name"

02.07.2013 10:46:40.917 Information Delivering to channel: In_JDBC

02.07.2013 10:46:40.917 Information MP: entering1

02.07.2013 10:46:40.923 Information MP: processing local module localejbs/CallSapAdapter

02.07.2013 10:46:40.924 Information JDBC Adapter Receiver processing started, required QoS BestEffort

Could this issue/delay be due to the number of mapping connections defined in PI (which basically controls the communication between ABAP and JAVA stack during mapping runtime)?

[Former Member]

Hi Prajeet,

Check the load on JDBC adapter at that time (for all JDBC interfaces at that time).

Try increasing the JDBC nodes with help of basis team so that JDBC threads availability increases, this requires system restart.

Thanks,

Pranav

[baskar_gopalakrishnan2]

>Could this issue/delay be due to the number of mapping connections defined in PI (which basically controls the communication between ABAP and JAVA stack during mapping runtime)?

The above log error does not show that completely delay due to mapping.  What do you do in mapping? Please elaborate your mapping functions. Can you please take that mapping data and test the same in other environments and see do you have any error?  Also you can pass the same data as query in toad or some client tool and check the results from DB.