How to enable sampling in jaeger-collector when using opentelemetry-jdk in java application?

[Bhav2308]

Hi All,

I am trying to implement jaeger remote based sampling.
Here, I am using a java application which uses opentelemetry sdk to send traces to jaeger-collector.
I am able to sample no traces at java application level using the following properties in application.properties:

otel.traces.sampler="parentbased_traceidratio"
otel.traces.sampler.arg="0.0"

But as i cannot restart my application multiple times because i need to change the sampling value multiple times, I was trying to achieve the same using jaeger-collector by passing sampler probabilistic in strategies.json:

SAMPLING_CONFIG_TYPE: file
    command: [
      "--sampling.strategies-file=/app/strategies.json",
      "--sampling.strategies-reload-interval=60s"
    ]  

and at jaeger-collector level it works very well, as in logs i am able to see this file is getting picked up:

2025-01-29 18:00:45 {"level":"info","ts":1738153845.2438943,"caller":"static/provider.go:149","msg":"Loading sampling strategies","filename":"/app/strategies.json"}
2025-01-29 18:00:45 {"level":"info","ts":1738153845.2483323,"caller":"static/provider.go:194","msg":"Updated sampling strategies:{\r\n  \"default_strategy\": {\r\n    \"type\": \"probabilistic\",\r\n    \"param\": 0.0\r\n  }\r\n}\r\n"}
2025-01-29 18:01:45 {"level":"info","ts":1738153905.238885,"caller":"static/provider.go:149","msg":"Loading sampling strategies","filename":"/app/strategies.json"}

But the problem arises, when I am still able to see spans being sent even though I have set the sampler value as 0.

2025-01-29 18:03:14 {"level":"debug","ts":1738153994.9348202,"caller":"app/span_processor.go:154","msg":"Span written to the storage by the collector","trace-id":"e6284f573414c73ebfb8226g30b325a2","span-id":"ccddfm530162028c"}

My guess is span coming from java application (opentelemetry sdk) are getting overwritten by jaeger-collector sampling strategy.
Is there a way, I can solve this?
What are the properties I need to configure at my application end to make this work?

[dosubot[bot]]

@Bhav2308 Hello! I'm here to help you with any bugs, questions, or contributions.

I found a similar closed issue that might be relevant to your situation. The issue discusses how Jaeger sampling configuration has no effect when using the AlwaysSample setting in the OpenTelemetry SDK, as backend sampling configuration only works with "remote" samplers. The OpenTelemetry Go SDK, for example, does not support Jaeger remote sampler out of the box, but there is a contrib module that does. This might be similar to your situation with the Java SDK ^[1].

Additionally, there are known compatibility issues between OpenTelemetry SDKs and the Jaeger collector regarding sampling strategies. The OpenTelemetry SDKs often lack necessary features such as sampler.type and sampler.param attributes, which are required for Jaeger's adaptive sampling to calculate proper sampling probabilities. This results in Jaeger adaptive sampling returning the default configuration for all services ^[2].

To address your issue, ensure that your OpenTelemetry SDK is configured to use remote sampling and that it points to the correct sampling service address. You might also want to check if the Java SDK supports remote sampling or if there is an extension available for it.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[yurishkuro]

You have to use jaeger remote sampler from contrib in your OTEL SDK

[Bhav2308]

@yurishkuro I am unable to find this io.opentelemetry.contrib.jaeger-remote-sampler or something similar in maven repo.
What is the correct way to include this?

[yurishkuro]

https://github.com/open-telemetry/opentelemetry-java/tree/main/sdk-extensions/jaeger-remote-sampler