[receiver/prometheus] speed up tests by setting skipOffsetting

[atoulme]

Fixes #32298

This PR modifies prometheusreceiver to allow to set the skipOffsetting option on its scrape option config. This option is private, so it can only be set by reflection.

This removes the random offset to start added to scraping prometheus metrics, so we may get faster CI builds.

[dashpole]

how much of a speed increase did it result in for you locally?

[atoulme]

I had 2 successive runs without errors (one failed on a different issue).

[atoulme]

3 runs without failures.

[atoulme]

how much of a speed increase did it result in for you locally?

I think I see overall a 20s reduction over the ~5minutes tests take. Given that we remove a random offset, it makes tests timing better. That said, I didn't manage to find why we wait 5s between starting the receiver and the first scrape.

[crobert-1]

That said, I didn't manage to find why we wait 5s between starting the receiver and the first scrape.

I didn't get a chance to confirm in this receiver, but go's ticker will first tick after your defined duration, by default. Users have to implement their own functionality to have a tick at start time. I assume that's what's going on here.

[crobert-1]

Any timing improvement here is helpful, as commented in the issue it looks like other-architecture builds are barely completing tests before the timeout 👍

Also, can you please add a reference to #32298 here? If the tests are consistently passing after this we'll be able to mark it resolved.

[atoulme]

Any timing improvement here is helpful, as commented in the issue it looks like other-architecture builds are barely completing tests before the timeout 👍

Yes, this is not a great improvement, but we can squeak by.

Also, can you please add a reference to #32298 here? If the tests are consistently passing after this we'll be able to mark it resolved.

Done

[atoulme]

4th run, no errors.

[atoulme]

5th run, no errors.

[crobert-1]

Since the goal here in my mind is to reduce timeout errors, I'd prefer this go into main sooner rather than later, even if we could theoretically continue to see errors intermittently. I'm going to add ready to merge unless someone has a different opinion. 👍