Ignore HTTP transactions by URL and header

[hegerchr]

Users should be able to provide a list of URLs and a list of request headers that should not be instrumented. If possible, they should be able to use central config to provide the lists. Both configurations should be dynamically applicable if possible.

[xrmx]

As discussed in the APM weeklies I'll try to implement this is as a RuleBasedSampler.

[xrmx]

PR for adding Composable Rule Based sampler upstream here open-telemetry/opentelemetry-python#4882

[xrmx]

Orthogonal to having a rule based sampler we should also check that all instrumentations pass enough attributes at the span creation for the sampler.

[xrmx]

Instrumentations to check for attributes passed to span creation:

• aiohttp-server, needs fixes: attributes added after span creation aiohttp-server: pass request attributes at span creation open-telemetry/opentelemetry-python-contrib#4118
• asgi, has user_agent and url.path
• django, has user_agent and url.path
• falcon, needs fixes: attribute added after span creation opentelemetry-instrumentation-falcon: pass request attributes at span creation open-telemetry/opentelemetry-python-contrib#4119
• fastapi, see asgi
• flask, has user and url.path
• pyramid, needs fixes: attributes added after span creation pyramid: pass request attributes at span creation open-telemetry/opentelemetry-python-contrib#4139
• starlette, see asgi
• tornado, needs fixes: attributes added after span creation tornado: pass request attributes at span creation open-telemetry/opentelemetry-python-contrib#4140
• wsgi, has user_agent and url.path

[xrmx]

Filed the issue with not enough data for sampling upstream here: open-telemetry/opentelemetry-python-contrib#4111

[xrmx]

All instrumentations PRs merged, rule based sampler still waiting for review.

[xrmx]

Rule based sampler merged, now I can draft an implementation of this in EDOT.

[xrmx]

Waiting for an upstream release in order to build central config support on top of that.

[xrmx]

Waiting for EDOT Java to complete the declarative config work and report back on the configuration side.

[xrmx]

We're going to wait for more declarative config work on the Python side as well.

[xrmx]

Declarative config work is in and should work fine for paths. The issue now is that most if not all the instrumentations does not pass the collected headers at span creation time and so are not visible to the sampler.

Filed open-telemetry/opentelemetry-python-contrib#4975