Add remote write 2 receivers compliance tests

[roidelapluie]

No description provided.

[roidelapluie]

cc @bwplotka FYI, WIP

[bwplotka]

Looks great! I think you gave us some ideas how to improve sender compliance (it tests too much IMO - it's literally scrape+export+Prometheus job/instance/staleness semantics, WDYT?)

What's your experience with the 2.0 protocol? Any feedback before we make it stable?

cc @pipiland2612 do you mind reviewing as well? (:

[krajorama]

Very nice indeed.

I understand it's not fully finished, my list of items needed:

• exemplars
• created timestamp
• different order of protobuf fields - this might be tricky, but for example I think Mimir right now would fail if the symbols table is after the timeseries

[roidelapluie]

Looks great! I think you gave us some ideas how to improve sender compliance (it tests too much IMO - it's literally scrape+export+Prometheus job/instance/staleness semantics, WDYT?)

I think testing RW1 is really difficult because of the lack of headers, out of necessity it is required to also test that the remote is an actual remote.

What's your experience with the 2.0 protocol? Any feedback before we make it stable?

cc @pipiland2612 do you mind reviewing as well? (:

My main experience is that there is quite a few blind spots in the doc.

e.g.:
Senders MAY add ;proto= parameter to the header's value to indicate the fully qualified name of the Protobuf Message that was used,, but it seems like for RW2 message this should be a MUST. But these are minor things.

[roidelapluie]

Pushed an update.

Results on cortex, mimir, prometheus:

https://roidelapluie.github.io/rwreport

[bwplotka]

Looks amazing! I'm encouraging @pipiland2612 to integrate and perhaps reuse index html table

[bwplotka]

e.g.:
Senders MAY add ;proto= parameter to the header's value to indicate the fully qualified name of the Protobuf Message that was used,, but it seems like for RW2 message this should be a MUST. But these are minor things.

Ups, will check, looks like a bug, thanks!

[roidelapluie]

Ready for review :-)

[roidelapluie]

Last test run: https://roidelapluie.github.io/rwreport/

[bwplotka]

Thanks!

[bwplotka]

BTW @roidelapluie I checked your feedback around:

e.g.:
Senders MAY add ;proto= parameter to the header's value to indicate the fully qualified name of the Protobuf Message that was used,, but it seems like for RW2 message this should be a MUST. But these are minor things.

I think the current form makes sense. It's MAY and no proto param means v1 for compatibility. That's why it's MAY.

Does it make sense?