This blog will give you tips for improved analysis in the ST22 short dump list. Questions that will be answered are:
What is this new button SAP correction notes doing there?
Detecting Z code in dump?
How to deal with RFC_NO_AUTHORITY dumps?
How to deal with CALL_FUNCTION_SINGLE_LOGIN_REJ dumps?
How to deal with TIME_OUT dumps?
How to deal with MESSAGE_TYPE_X dumps from program SAPLOLEA?
How to deal with TSV_TNEW_PAGE_ALLOC_FAILED dumps?
How to deal with CALL_FUNCTION_BACK_REJECTED dumps?
How to authorize ST22 details?
How can I find dumps specific to SAP HANA?
The tips below assume you have not yet any clue on how the dump is created. If you know how the dump is created, but cannot analyze from the dump: please read the blog on the ANST tool: automated notes search tool.
Detecting Z code in a dump is normally easy if it is a Z program. Some dumps you can have due to the fact that Z code is there in a user-exit, which again is calling SAP code. This dump will appear as looking 100% standard SAP, but when you scroll down in the Call Stack you will see Z code: Before raising OSS message to SAP: make sure the call stack does not contain custom Z code.
The RFC_NO_AUTHORITY is special kind of dump and typically looks like this: First thing to get from the dump is the user ID and the calling system (is it an internal call or call from different system). And if the user ID is a human user or system user. Second thing to determine is: is this a valid call or not a valid call? In case of valid call, look in the dump which authorization is missing and what needs to be added. If the addition is done: do keep an eye on the dumps, since a new dump might come for a different new authorization object. In case of an invalid call, you need to determine how the call was initiated and take action to avoid the initiation. This is not always a simple job. Why is checking this dump important? Complete business flows might be disrupted if this happens. It is hard to detect for the end users what is going on. It will take them time to raise an incident and for functional people to determine what is going on. This way a lot of valuable time can be lost. What can also happen: people try to connect via RFC methods to read data. This will give lot of dumps which are hard to follow up. If you get too many of these dumps and you can’t solve them, you can switch parameter rfc/signon_error_log to value -1. Then the dumps are no longer there in ST22, but in stead moved to SM21 system log with less detail. If you need to have the details again, switch the parameter again (it is dynamic). Background on the parameter rfc/signon_error_log can be found in OSS note 402639 – Meaningful error message texts (RFC/Workplace/EBP).
A bit similar to the above dump is the CALL_FUNCTION_SINGLE_LONG_REJ dump. Here a user tries to login via RFC to the SAP system, from a different SAP system, or from a JCO based connector. Again: first determine if the call is valid or not. If not valid, determine the calling source (can be hard!). If it is a valid call, scroll down in the details section for this dump and look for the part below: There are two codes: T-RC code and the L-RC code. Check both the codes. In this case above the user ID validity was no longer ok. Depending on the codes different solution needs to be applied. Why is checking this dump important? Complete business flows might be disrupted if this happens to system user. If it happens to single user he might get grumpy. It is hard to find for the end users what is going on. It will take them time to raise an incident and for functional people to determine what is going on. This way a lot of valuable time can be lost.
If an online query takes longer than the timing set in parameter rdisp/max_wprun_time a TIME_OUT dump will happen. By default and best practice, this time out parameter is set to 10 minutes. This is also the case in most system. This dump will look like: If you scroll down (or click in the left section) to the User and Transaction section, you can see the ID of the user who caused this and the transaction. First reaction of the average basis person is: call/mail the user and ask him to run this in batch mode. This is indeed one of the solutions. Alternative potential solutions:
Analyze with the end-user if he can fill out more selection criteria (hence reducing the time needed to select the data)
Analyze with the end-user if he can run the report in multiple smaller sets
Check if there are known performance OSS notes for the transaction the user is running (the root cause might simply be an SAP bug)
Check if the database statistics of the tables queried is up to date
In some cases both the selection criteria are ok, and the output of the list in batch only give a few results: in this case the creation of special index might be the solution. This can happen in case of check reports that look for business exceptions.
Why is checking this dump important? Users tend to get very frustrated by the system if they hit this dump. They have to wait 10 minutes and have no result. Sometimes you see this dump a couple of times in a row. Imagine yourself being the user with a boss demanding report which crashes after 10 minutes…
MESSAGE_TYPE_X dumps from program SAPLOLEA
The MESSAGE_TYPE_X can be pointing to very serious issue. But the ones generated by program SAPLOLEA point towards one type: the SAP GUI server interaction. This dump typically look like this: a main dump MESSAGE_TYPE_X and calling program SAPLOLEA. This dump can have 3 main root causes:
Issue in ABAP code (hit the SAP correction notes button to search for solutions)
Issue in local SAP gui installation of the end user
Issue in the SAP kernel
If you see many dumps with the same user ID: this typically points towards an old local SAP gui installation. Solution is to update the local SAP GUI for that user to the latest version that is supported in your company. In rare cases the SAP kernel causes these kind of dumps. These are hard to find and detect. The only remedy here is to update the kernel at regular intervals. To find which users use which SAP GUI version: go to transaction SM04 and add the field SAP GUI version: From ABAP code: use function module TH_USER_LIST to get list of sessions. The GUI version is in the field GUIVERSION of output table USRLIST. For more background on SAP GUI patching read this dedicated blog.
These dumps are caused by missing callback positive listing. See OSS note 2981184 – What to do in case of CALL_FUNCTION_BACK_REJECTED short dump. The solution is to add the function module to the positive list in RFC. In no way reduce the RFC security by lowering the RFC callback security parameter rfc/callback_security_method. Read this blog on how to hack using callback RFC, and why not to lower the security.
Coding and table generation dumps
Dumps can happen due to coding and tables not generated properly. When it happens during transport import, it is normal. If it persists after the import, you need to act. Best practice notes:
This blog will explain one of the most useful new tools from SAP when having to find bugs in standard SAP coding. The ANST (automated notes search tool) is not receiving the recognition that is should get. In usability it is same ease as the SNOTE tool.
If you love SNOTE you will also love the ANST tool! Just try it out.
Questions that will be answered in this blog:
What is the ANST automated notes search tool?
How does is work?
Why should I always use this tool before submitting an incident to SAP?
ANST (advanced notes search tool)
The ANST tool can help you in:
Quickly finding OSS notes for your issue
Check if you Z code is causing the issue or dump, or it is a standard SAP issue
OSS note 1818192 is the ANST FAQ note which also has the minimum version. This note also has an extensive explanation. The how to use below is just a summary.
OSS note 2605555 also contains an excellent PDF inside as attachment, that gives a step by step manual.
How to use the ANST tool?
Start transaction code ANST.
If you launch it for first time you might get an error "ANST001 Fatal Error. Customizing table is not filled". If this is the case follow the solution steps in OSS note 1909768.
In the transaction code box key in the transaction where you have the issue. As example we will use transaction code S_BCE_68001417 (search for authorizations by complex criteria). The user admin is complaining about an incorrect number of selected authorizations that are shown in that transaction.
So key in the transaction code and description (you can keep it same).
Now press execute: the transaction will be called. In the authorization object screen fill out S_DEVELOP and execute again to get the results:
Now leave the transaction recording.
In the left bottom of the screen you can see the recording being written into the trace file:
Depending on the complexity and amount of screens you have passed this can take up to 1 to 10 minutes.
The result is shown after the trace file. The result is sorted per SAP module. If you open the details, you can also see the exact program blocks that were hit during the recording.
Now you can select the modules (if wanted specific code blocks) where you thinks is the issue. After selection hit the Note Search button. The SAP system will now connect to SAP service marketplace and look for the most recent notes for your version, which have not yet been implemented.
The middle note seems to be very relevant. From this screen you can can already link to the note (click on note number) and start download to SNOTE already.
Tips on the selection of the components:
1. Never select more than 1000 components: ANST will reject this
2. The less components you select the faster you get results, and shorter list of potential notes as well
3. If you want you can later retrieve the recording and make a different search on different components: no need to re-record
4. Most of the times you can ignore the basis and cross application and basis notes
5. Run the recording and the result together with your functional consultant: he can help filter the components and select useful notes
Changing settings for maximum amount of notes
Especially in the basis or core ABAP area you will notice that ANST cannot read more than 1000 notes at once. This is a default setting you can easily change. In the main ANST screen hit the Settings button and scroll to the right to increase the maximum notes number:
Using ANST to analyze short dumps
The ANST tool can be used as well to analyze short dumps. Just start the ANST tool and run the steps including the step where the dump occurs. After the dump the ANST tool will trace the modules including the point where the dump occurs.
Make sure OSS note 2535278 is applied: this contains bug fix for the short dump case.
Checking for customer code issues
After the trace file is generated and you have searched for OSS notes, it can be there is still an issue caused by your own customer code. To exclude this (or to check it anyhow), you have to use the button Customer Code from the trace result screen with all the components. Be a bit patient while the tool is scanning for modifications, user-exits, BADI implementations and enhancement spots it came across in the recording.
If you want to analyze implicit and explicit enhancements as well with ANST you must apply OSS note 2408785 first.
ANST clickable demo
SAP has made a nice clickable demo to show you how it works: link to demo.
Use of ANST tool before submitting incident to SAP
Even if the ANST tool does not help you search for the correct OSS note for your issue, the ANST tool can help you in speed up of the incident solution for SAP.
If you want to report the issue to SAP as an incident download the ANST trace file. If you report the incident mention:
ANST tool is used and add the recording
Add list of already implemented OSS notes
You already checked for customer code
With this information the first line processor will have a quick job assigning the incident to the real issue solvers in Walldorf. This will save you valuable time, since the first line normally come with simple list of notes, or also run the ANST tool themselves, and then come with obvious notes.
Increasing the maximum number of objects limit
If you are using the ANST tool on a transaction with many objects (for example ME21n purchase order), you will notice that you cannot search for more than 1000 objects at the same time. Then you have to open subsection and select subtree and run it more than once with different selections. But sometimes one node really expands into more than 1000 objects. In this case, you best increase the maximum object limit. In ANST start screen choose the Settings button can increase the Max.Object counter on the far right of the settings (scrolling required)
Needless to say, more objects do take more time to analyze. But it is worth the wait.