How is a constant field throwing error with missing

[Former Member]

Hi Experts,

we are facing a weird case in production PI 7.1:

The message is sending from PI to ECC via IDoc, and sometimes(not all times) we got error saying a specific node is missing from the IDoc adapter as below.

Actually this node is mapped with constant in Message Mapping, so it shouldn't be missing for the IDoc adapter to process.

As this IDoc is a standard IDoc, which is showing last update was some day in 1998.. and PI imported this IDoc in 2010, so I don't think it is related to metadata in PI(idx2), actually even tried to delete the metadata but still failing afterwards.

However, the most weird thing is the failed message can be reprocessed successfully without doing any thing-- why it was failed at the beginning?

Thanks for any idea shared!

[former_member184720 · Accepted solution]

Hi Hailong - The same problem resolved by applying the notes #1574923 & 1609818 and i believe it's the same issue with you too.(as per your screenshot, i see that you have changed the IDOC structure from 1 to 0..unbounded)

Source -

[markangelo_dihiansan]

Hi Hailong,

Probably a mapping bug, please remove the constant mapping for E1DKU1 since the occurrence of this node is 1..1 it will always be present in the output.

Regards,

Mark

[Former Member]

Hello Mark,

I'm afraid this is not true. If you don't map a mandatory field on target side the mapping will throw an error. You need to feed mandatory fields with something, otherwise the mapping will fail.

If it only fails for some messages and reprocessing solves the issue, I assume there is a bug somewhere else, but not in the mapping.

Regards,

Jörg

[markangelo_dihiansan]

Hi Jörg,

Thanks for the comment. I meant to say that that node does not require a mapping since its occurrence is 1..1, fields that have 1..1 or 1..unbounded do require a mapping. For example:

In the message type above, Val1 and R1Val1 are both mandatory fields and they need a mapping (hence they are red). Val2 and R1Val2 are grey, means they are not mapped. But for MT_Sample2 (Root) and Row1, they are already in green means mapping is complete even if they are not really mapped at all (see yellow for Row1). Executing it, will produce just the root node and row 1.

That is the reason I asked him to remove the constant mapping for that field

Regards,

Mark

[Former Member]

Hi Mark,

I just confirmed that in my own system, you're right. This must have been introduced some time ago, I wasn't aware of that change. Sorry for calling a mistake where there was none!

Regards,

Jörg

[markangelo_dihiansan]

No worries Jorg, it is okay!

[Former Member]

Thanks Mark, but do we have evidence if we mapped a constant to such mandatory field will cause error?-- Although this mapping is not required, it shouldn't populate error either...

This a issue in production, which means I can't directly make the change and check the results, so I need to find a solid proof this will causing issue so that they may agree to make such changes.

As suggested by Gampa, I am also checking the 2 SAP notes mentioned.

Appreciated if you have any comment for that SAP notes!

[markangelo_dihiansan]

Hi Hailong,

Unfortunately, I do not have proof. The best that you can do is to replicate this issue in your dev system. The SAP notes though are worth to try.

Regards,

Mark

[Harish]

Hi Hailong,

If the error not occure always then the issue is with PI server level. Are you getting error with all message during that time?

regards,

Harish

[Former Member]

Thanks for the response, Harish,

Some messages were succeeded while some were failed at the beginning, and we have to reprocess them, and all of them can be reprocessed successfully.

The issue happens nearly every day(1-3 times).

[Former Member]

Hi Hailong

Generally this error comes when the meta data structure is not updated in the PI side.

Can you please confirm that it is not happening for all idocs ? and you are seeing it only for specific idocs.

Thanks

RD