Refactor tail-sampling processor - Refactor policies

[jpkrohling]

• [ ] Create a new pkg/samplingpolicy with the interface
• [ ] Migrate one policy (OTTL) to be an extension
• [ ] Add a new policy to the tail-sampling processor called extensions, which would load the policy extensions
• [ ] Evaluate which other policies need to be migrated

Quite a few of the policies can be replaced by OTTL, while others could be replaced by OTTL if new features are added to it. For instance, the latency policy can't be done yet with OTTL.

[jpkrohling]

@TylerHelmuth, do you think it would be feasible to have a duration func for OTTL, to calculate the latency of a trace?

[github-actions[bot]]

Pinging code owners for processor/tailsampling: @jpkrohling. See Adding Labels via Comments if you do not have permissions to add labels yourself.

[TylerHelmuth]

@jpkrohling you can use OTTL's built-in arithmetic to get a span duration, but doing any function for the entire trace would be tricky. OTTL currently can work on a Span but not a Trace as a whole.

[github-actions[bot]]

This issue has been inactive for 60 days. It will be closed in 60 days if there is no activity. To ping code owners by adding a component label, see Adding Labels via Comments, or if you are unsure of which component this issue relates to, please ping @open-telemetry/collector-contrib-triagers. If this issue is still relevant, please ping the code owners or leave a comment explaining why it is still relevant. Otherwise, please close it.

Pinging code owners:

• processor/tailsampling: @jpkrohling

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[alec-w]

Not sure if this is the right place to ask this but it does tie into the tail based sampling policies...

When making use of the ADOT collector in EKS we're using the remote sampler (which essentially fetches the sampling rules from AWS XRay).

This sampling decision is still head based though.

We'd like to use tail based sampling so we can always record traces for long-running requests (which the current tail-based sampling processor would allow us to do) but use the XRay sampling rules for all other requests (which it wouldn't).

Would the only way to accomplish this be the addition of a new processor? It seems that such a processor would in many ways end up being very similar to this one - so wonder if our thinking is heading in the right direction...

[jpkrohling]

This would likely be a new policy. Under the new architecture after the refactoring, it would be an extension.