Orion-LD restarts when a large payload is given to a Property value

[ttutuncu]

Hi,

We are having a problem causing Orion-LD to restart when we try inserting the example entity attached through ngsi-ld api.

requestPayload.txt

I believe the problem is not because of the structure but because of the length of the value of the Property. We can successfuly insert this model when we have fewer values in the property, without causing a restart.
Strangely, while it causes a restart, the data is being written to MongoDB.

Tested Versions:

• Orion-LD: v1.4.0
• MongoDB: v3.6.23-13.0

[kzangeli]

So, I took a quick look at the "entity with the huge attribute value" and ...
Well, let's just say, this is not how you should do it, to say the least !

You might as well stop using the broker and just put that entire entity in a file in the file system and forget about queries or anything and just retrieve it as is. E.g., you can't query over "buildingNo" or "streetName" or anything at all. You'll always get that entire entity back with ALL the information in it.
Doesn't make very much sense to me.

A data model that would make more sense (to me) would be to have every building as separate entities and this "mega attribute" ("buildings") having a Relationship-array to all those entities.

We (FIWARE) have the smart data model initiative in place to help with this kind of problems. I can point you in the right direction to get information about all that.

That said, the broker must never ever crash, no matter what you feed it with, so, we'll start looking into this crash right away.

[kzangeli]

I implemented a functest creating the entity in question, expecting a crash, but it worked just fine.
You see the PR referenced above (#1474 )

So, as I can't reproduce, I'm gonna need more input.
E.g., complete set of headers in the entity creation request (POST /entities), all env-vars/cli-params you start the broker with. Stuff like that.
Hopefully with all that info the issue can be reproduced. If not, it's not possible to fix anything and I'd suggest you look on your side.
We could enable all traces, to start with, and see if that sheds any light on the problem.

It still might fail in github actions, we'll know in about 30-40 minutes ...
Well, it failed, but, for other reasons. I fixed it and all OK.
So, more input needed. Right now, this seems like a non-issue