Dumps & Changes not saved

[perlundh]

Hello community,

Before creating a OSS-ticket... Is there any common problem with Personas giving these errors?

We have big problems with Personas that I have not encountered before. We get a lot of errors and strange behaviours when Saving in the Flavor Edit.

• 20% of the time, we get an error in the web browser (500 internal server error) and SM21. Have to re-do everything. Sometimes during the day it almost happens 80% of the times you save, sometimes during the day it's 20%. We have tried to find a common thread but have not found any.
• 50% of the time when you Save, the changes are not actually saved and when you exit the editor - your changes are gone. Very frustrating.

We have tried to find a reason, updated all OSS notes in the health check, clear flavor cache, clear mime cache, followed note 2412840, 3071141, 15766 to update kernels etc...

We are mainly doing changes to the EWM RF-transaction but the problem is also there for other transactions..

Signal c00 received by the operating system is the most common error in SM21 but we have also seen things like Runtime error "TABLE_INVALID_INDEX" occured.

[tamas_hoznek · Accepted solution]

Hi Per,

These errors are definitely not expected and not acceptable in any properly updated and supported Personas environment. They seem to point to something more of a system (aka basis) problem rather than anything Personas-related. If there was recently some change (such as an upgrade or basis / DB maintenance), this could result in such a situation, but it is difficult to tell what the culprit may be without analysis.

I would suggest creating an incident, but it is equally important to be able to safely reproduce such errors, otherwise investigation becomes difficult if the problem appears in a random fashion. In the ticket, you can describe the steps to recreate the problem.

[perlundh · Accepted solution]

Thanks Tamas.

Yeah, I have had a lot of support from our basis team but we can't pinpoint or solve the issue..

I will create an OSS-note after summer vacations so we have all people onboard to work with the incident updates 🙂

[perlundh]

I have now created a ticket tamas.hoznek.

As you wrote, I just hope that the issue can be reproduced as it is not 100% "linear".. They need to "click around" in the flavor editor for a couple of minutes before the dumps and errors starts to show. Hopefully we can find a fix otherwise we need to skip using Personas which would be a shame!

[tamas_hoznek]

An important question I didn't ask earlier:

Is there any difference in behavior when you are editing flavor(s) in SAP GUI for HTML vs. Slipstream?

[perlundh]

Good questions. I think I need to update the ticket with this information.

First of all, it's only when editing flavors - not when using. It's however sometimes veeery slow when we run our current primary flavor which is the RF Framework in EWM. That might however be a separate topic.

We have only used Slipstream Engine so far.

I have tested with GUI for HTML this morning and I am not able to reproduce it. On the other hand, I'm not able to reproduce in Slipstream either this morning (I got errors every 3-5 min on Friday for the same flavor in the same system). I'm now doing this in our Sandbox environment. I get a feeling that it's some type of buffer-thing because in our development system, our flavor is much bigger (we have changed all screens in RF) and there we get the errors constantly and also that changes are not saved even though we save and exit.

Sometimes, you can do Save+Exit and it looks OK but then 2 days later - it's back to original state (save have been reversed). It's reeeeeeeally frustrating and it scares me that it's not predictable to reproduce.

[tamas_hoznek]

Hi Per,

I saw that your incident is in process by our support team. I linked this thread to it, so they are aware of the additional info provided here.

Indeed, it sounds that something of the sort you are suspecting may be the culprit here. Since the problem is most apparent with EWM RF flavors, I was wondering if certain actions could help matters to avoid the errors when saving. EWM RF flavors typically contain a lot of augments, which makes the change history huge. I wonder if compressing the undo history for such a problematic flavor makes a difference. Clear the flavor cache again after doing this and check if the situation improves.

Also, depending on the kind of changes the flavor contains, sometimes it makes sense to consolidate as many formatting changes as technically possible into a theme, instead of adding them to a flavor screen by screen or control by control. Did you try to take maximum advantage of a theme for such changes?

[perlundh]

Good ideas Tamas!

I have tested to "Compress undo history" and also "flavor cache" but maybe not after each other. Will try that now.

And yes, I understand the RF might be the most complex "transaction" as it is 500 screens, flexible logic etc.. We are probably one of the companies doing Personas on RF the most, using it instead of the "graphical screen editor" to move fields etc.

The only "general change" suitable for theme would be the font size that we have increased in each screen. The reason to do it in each flavor was however to have the option of changing this in combination with an adaptive flavors for different screen sizes/devices. As a general rule, I try to avoid changing "theme"-related objects as I want the flavor to work with future standard themes.

Worth noting - these issues also appear in "normal transactions". I asked my basis resource to test editing VL32N and after a couple of changes and minutes he got the dump/error with internal server error.

However, it's only in our big RF flavor that we see the problem with "Save + Exit" having no effect or being reversed a couple of days later.. This problem worries me a lot!

[perlundh]

I did "Compress Undo History" and "Clear Flavor Cache" now in our development system.

I have now been able to run Personas and do change for an hour without getting kicked out or save-issues. Usually, you get it each 5-10 min. I will continue to test during the day.

However, I still have the feeling that something is wrong on a basis/infra-level and as mentioned and that the issue will come back. We have updated kernels etc etc according to notes we found.

Edit.... and exactly when I wrote this the "Save"-errors started. Just like a buffer of something hade filled up during my "good hour". I added a video clip of this in the Support Ticket now. Each time you Save and Exit - something random is reverted back to "original state" and only some changes are saved.. Very frustrating!

Edit 2... I did Compress Undo + Flavor Cache again and after that it's quicker to use the editor but still the random "Save"-errors..