[receiver/k8sobjects] ensure the k8s.namespace.name attribute is set for objects retrieved using the watch mode

[bacherfl]

Description

This PR ensures that the structure of log records generated by the k8sobjects receiver is the same, regardless of the mode (watch or pull) being used. This also solves the issue of the k8s.namespace.name attribute not being set for objects retrieved with watch mode.

Link to tracking issue

Fixes #36352

Testing

Added unit tests and adapted e2e tests

[tetianakravchenko]

FYI: there is an issue regarding reflecting of k8s events to the events API definition - #35857 (comment), maybe it makes sense to include event.type to the events api convention

[ChrsMark]

Thank's @bacherfl for working on this!

I agree that both pull and watch modes should populate the k8s.namespace.name resource attribute as it was highlighted at #36352.

However, I'm not sure if the existing structure should change, object.metadata.* -> metadata.*.

Isn't this a breaking change? Should we ensure if we really want this or not?
Otherwise, if we can fix the namespace issue without introducing a generic breaking change that would be optimal I think.

Could we clarify the above questions?

[bacherfl]

Thank's @bacherfl for working on this!

I agree that both pull and watch modes should populate the k8s.namespace.name resource attribute as it was highlighted at #36352.

However, I'm not sure if the existing structure should change, object.metadata.* -> metadata.*.

Isn't this a breaking change? Should we ensure if we really want this or not? Otherwise, if we can fix the namespace issue without introducing a generic breaking change that would be optimal I think.

Could we clarify the above questions?

Thanks @ChrsMark for the feedback! You're right, changing the structure reported by the watch mode to be the same as with pull would be a breaking change, so I can also adapt the PR to specifically fix the namespace being set, if that is more preferable. In this case I would also add a note in the readme that the structure of the emitted log entries might differ, based on which mode is being used for the receiver

[ChrsMark]

Thanks @ChrsMark for the feedback! You're right, changing the structure reported by the watch mode to be the same as with pull would be a breaking change, so I can also adapt the PR to specifically fix the namespace being set, if that is more preferable. In this case I would also add a note in the readme that the structure of the emitted log entries might differ, based on which mode is being used for the receiver

Cool! I think it's just fine fixing the namespace issue here and possibly file another issue to evaluate the "inconsistency" between the 2 modes. (in case we decide to change this, it should probably be done with a feature gate etc).

To my mind there should be no difference in the 2 modes from users' perspective but maybe I miss the details here behind this decision.

[bacherfl]

Thanks @ChrsMark for the feedback! You're right, changing the structure reported by the watch mode to be the same as with pull would be a breaking change, so I can also adapt the PR to specifically fix the namespace being set, if that is more preferable. In this case I would also add a note in the readme that the structure of the emitted log entries might differ, based on which mode is being used for the receiver

Cool! I think it's just fine fixing the namespace issue here and possibly file another issue to evaluate the "inconsistency" between the 2 modes. (in case we decide to change this, it should probably be done with a feature gate etc).

To my mind there should be no difference in the 2 modes from users' perspective but maybe I miss the details here behind this decision.

alright, then I'll adapt the current PR accordingly and will file an issue regarding the two different data structures - will let you know when this one is ready for another review :)

[ChrsMark]

lgtm