Dev-docs: Add HTTP Tracer explanation doc

[prateek041]

No description provided.

[CLAassistant]

All committers have signed the CLA.

[marctc]

hey @prateek041, is this still a draft? do you want keep adding stuff or we can review it?

[prateek041]

Hey @marctc so I raised this MR a while back because I was assessing if the project wanted documentation like this on the project. Specifically in the DevDocs sections.

If the project has a need for it, there is tons of more content I can write.

[marctc]

Hey @prateek041, that sounds amazing. We would like to have such documentation like this. Feel free to put as ready for review and we can take a look.

[rafaelroquetto]

The HTTP tracer has been removed - the job is now done in https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation/blob/main/bpf/tpinjector/tpinjector.c

[prateek041]

@rafaelroquetto thanks for pointing that out, but the idea behind, I.e. attaching the trace-parent header is still looks the same.

[rafaelroquetto]

@prateek041 indeed - however, it may be worth updating the documentation to reflect the current approach which relies on skops and sk_msg programs on egress to inject the traceparent (and not TC) - for reference, this is where it happens: https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation/blob/main/bpf/tpinjector/tpinjector.c

This class of programs no longer requires us to manually compute the TCP checksum or deal with the sequence numbers, as they operate at a higher level.

We still have a TC egress program for injecting the traceparent on the IP header in case the above fails (e.g. SSL), and another one on ingress to extract the options from the IP header - the code is here: https://github.com/open-telemetry/opentelemetry-ebpf-instrumentation/blob/main/bpf/tctracer/tctracer.c

[prateek041]

Thanks for the detailed response @rafaelroquetto I will definitely update this over the weekend.
I'm actually glad that we don't need to manually calculate the TCP headers anymore.