Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add a more flexible sampling strategy data model #107

Open
wants to merge 3 commits into
base: main
Choose a base branch
from

Conversation

garrettlish
Copy link

Which problem is this PR solving?

Description of the changes

  • Add a more flexible sampling strategy data model, it enables support for more fine-grained control, such as adjusting sampling probabilities based on specific tag key-value pairs (see sampling.proto).
  • For instance, in a real-world scenario, we might want to enforce sampling for a particular user session by customizing sampling probabilities based on specific tag key-value pairs.

How was this change tested?

  • Tested in the upcoming PR.

Signed-off-by: xiancli <xiancheng.li@ebay.com>
@yurishkuro
Copy link
Member

This feels too simplistic to me. Compare with open-telemetry/oteps#250

Signed-off-by: xiancli <xiancheng.li@ebay.com>
@garrettlish
Copy link
Author

This feels too simplistic to me. Compare with open-telemetry/oteps#250

Good idea. How about introducing an "any" or "conjunction" option, along with adding a dimension matching strategy, similar to the changes made in this commit - 171bccd? Any thoughts, please advise!

@yurishkuro
Copy link
Member

I would recommend starting with some use cases and showing a yaml-equivalent of the config for those, then we can translate it to the IDL. I think your proposal is still too rigid. For example, here's the DLS that Gemini recommends:

message SamplingRule {
  // Logical expression combining conditions
  LogicalExpression expression = 1;

  // Sampling strategy
  oneof sampling_strategy {
    FixedRateSampling fixed_rate_sampling = 2;
    ProbabilisticSampling probabilistic_sampling = 3;
    // Add other sampling strategies as needed
  }
}

message LogicalExpression {
  oneof expression {
    Condition condition = 1;
    AndExpression and_expression = 2;
    OrExpression or_expression = 3;
    NotExpression not_expression = 4;
  }
}

message AndExpression {
  repeated LogicalExpression expressions = 1;
}

message OrExpression {
  repeated LogicalExpression expressions = 1;
}

message NotExpression {
  LogicalExpression expression = 1;
}

example rule

message SamplingRule {
  expression {
    and_expression {
      expressions {
        condition {
          attribute: "http.status_code"
          operator: GREATER_THAN
          int_value: 500
        }
      }
      expressions {
        condition {
          attribute: "service.name"
          operator: EQUALS
          string_value: "my_service"
        }
      }
    }
  }
  fixed_rate_sampling {
    rate: 0.5
  }
}

The challenge when dealing with expressions and boolean logic is in validating that the rules in a strategy are disjoint, as otherwise the stratas are going to overlap and proper stratified sampling probabilities cannot be determined (although we probably can deal with that by assuming a priority order of the rules where earlier rule wins, in which case even if two rules overlap in the search space prioritizing one of them will effectively make them disjoint).

Signed-off-by: xiancli <xiancheng.li@ebay.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants