BW system ran out of "session" => transaction SMGW; how to track down "resource" issues?

[behlau_carlos]

Hi SAP Experts,

we had on our SAP BW system the following situation:

We are having a week daily process chains, that is uploading data from our ECC system.

Once in a month, we are uploading from our APO/SCM system data from our BW system.

Parallel we are having also SAP PreCalculation Server (with three PreCalc instances) instances running. The PreCalc Server is running daily.

The landscape in short: (DB server = AIX/Oracle, App-Server = Linux, SAP System=7.01; PreCalc Server = Win2008/64)

This week when all was running at the same time, we faced problems, that log-on to SAP System via ABAP took a long time, same BEx Reports (some failed with Run Time errors).

From the landscape infrastructure we couldn’t detect any resource issues (we are having two application servers connected to the BW landscape).

What we saw from the system transactions (SM66, SM51, SM21, ST02, ST04, ST06N) the issue did not seem to be resource based, by network, database, amount of work processes etc.

But we noticed in transaction SMGW, that it seemed not possible for external programs, like a BEx reports or the PreCalc Server to connect with new session to the BW ABAP landscape.

The trace showed in SMGW:

*****************************************************************************

*

*  LOCATION    SAP-Gateway on host server / sapgw##

*  ERROR       Conversation 01792797 not found

*

*  TIME        Tue Jul 22 14:13:03 2014

*  RELEASE     720

*  COMPONENT   SAP-Gateway

*  VERSION     2

*  RC          728

*  MODULE      gwxxconn.c

*  LINE        971

*  COUNTER     75240

*

*****************************************************************************

In system log, database alert log, we could not find any dead locks reported or anything reported, that explained why, via the SMGW is was not possible to connect with new sessions.

We restarted the PreCalc Server and after clarifying with application team of APO/SCM we stop the process chain and than the system was back to “normal” operation.

Now the question is, if you faced similar issues and what the solution, or how you where able to track down the issue?

Thanks for your ideas.

Best regards

Carlos Behlau

[alwina_enns · Accepted solution]

Is the affected system on a higher patch level as described in the note:

2006469 - GW: Gateway does not accept any new connections

[behlau_carlos · Accepted solution]

Hi Alwina,

that could be really the cause of the problem.

We are running on Kernel 600 and the solution is delivered in kernel patch 616.

I am going to organize a kernel upgrade.

That might take some time, as need first to upgrade development and test system.

As soon as the kernel update has been done, I am updating this post with the results.

Best regards

Carlos

[alwina_enns · Accepted solution]

Hello Carlos,

in case of bottlenecks you should see messages about it in dev_rd, in ST03N you see only response times, but not the number of connections. Screenshots about active connections in SMGW are helpful, which status have the connections. To analyse if enough memory is configured for gateway you can activate gateway statistics under "SMGW --> Goto --> Expert functions --> Statistics --> Activate" during issue occurs. If you will see "red" connections in SMGW please take screenshots about details of such connections ("choose detail" button, details when you double click on a connection, details when you click on conversation ID). It is helpful to check also the communication partner, which errors are reported on the other side.

Regards,
Alwina

[alwina_enns]

Hello Carlos,

what do you mean with

"But we noticed in transaction SMGW, that it seemed not possible for external programs, like a BEx reports or the PreCalc Server to connect with new session to the BW ABAP landscape."

How many active connections you have seen in SMGW - was it near the maximum?

Regards,
Alwina

[behlau_carlos]

Hi Alwina,

thank you for your help.

What I was able to is in SMGW, because the system was very slow, that some connection where "red".

I forgot to capture screenshots and now I am not able to recall the the status text.

It is there a way to find out, if we reach the "max" connection numbers, like parameter gw/max_conn 500, via ST03N or via SMGW or via some logs?

I checked the SMGW trace, but I am not able to identify via the trace in SMGW, if the failures happened now, because max. values of parameters have been reached or what the cause was/is.

What I meant:

All external programs, like SAP BW BEx Analyzer (Excel Reporting tool for BW) are connecting via RFC calls. Also SAP Pre-Calculation Server is using RFC calls to generate the reports and send them via E-mail.

Usually, I can see these connections in SMGW (goto => logon clients).

There I didn't noticed that we reach 500, when we had the problem.

Best regards

Carlos