Response success status validated incorrectly

[slyapustin]

Describe your environment

OS: (e.g, Ubuntu)
Python version: (e.g., Python 3.8.10)
SDK version: (e.g., 1.25.0)
API version: (e.g., 1.25.0)

What happened?

There are some places, where success status of the request validated based on only 200 and 202 status code:

opentelemetry-python/exporter/opentelemetry-exporter-zipkin-json/src/opentelemetry/exporter/zipkin/json/__init__.py

Line 94 in 6812da2

REQUESTS_SUCCESS_STATUS_CODES = (200, 202)

opentelemetry-python/exporter/opentelemetry-exporter-zipkin-proto-http/src/opentelemetry/exporter/zipkin/proto/http/__init__.py

Line 91 in 6812da2

REQUESTS_SUCCESS_STATUS_CODES = (200, 202)

Steps to Reproduce

Receive the HTTP 204 succes status code

Expected Result

204 status to be treated as a success status

Actual Result

204 status code treated as an error

Additional context

Relevant issue (fixed) about the same behavior in another module: #3621

Would you like to implement a fix?

None

[codefromthecrypt]

zipkin doesn't define its HTTP POST status as returning 204, rather 202 as the code already does.

https://github.com/openzipkin/zipkin-api/blob/master/zipkin-api.yaml#L83

What product is returning 204? Can you raise an issue with them to switch it?

[stfines-clgx]

I get it when using Arize Phoenix and Elastic Search. Maybe somebody could convince arize, but I doubt that one could convince elastic.

[codefromthecrypt]

hmm the collector codebase returns correct HTTP status. I'll ask my colleagues at Elastic and also my friend at Arize

https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/70ab6e9b8ae96e4b7c59188672b1de0c93bb3e7d/receiver/zipkinreceiver/trace_receiver.go#L247

[stfines-clgx]

I'm getting mercilessly spammed with 204's hitting the Arize open-telemetry http endpoint when I do local development -
DEBUG:urllib3.connectionpool:http://localhost:6006 "POST /v1/traces HTTP/1.1" 204 0 ERROR:opentelemetry.exporter.otlp.proto.http.trace_exporter:Failed to export batch code: 204, reason:

[codefromthecrypt]

zipkin's POST endpoint is /api/v1/spans, and that's what the collector correctly binds as well.

Seeing /v1/traces (OTLP) makes me think possibly there's a mismatch on exporter/receiver.

p.s. By default, zipkin itself doesn't bind the OTLP endpoint /v1/traces, though there is a different docker image that supports that recently

[stfines-clgx]

well, I'm using Arize Phoenix, not zipkin so I'm not sure where that comes in to play here.

[codefromthecrypt]

@stfines-clgx so the issue as opened was a complaint about the code in zipkin exporter (not OTLP), and if it should accept HTTP 204 (which is not something defined by zipkin as it defines it uses 202).

I think your concern is related to 204 code, but not related to zipkin defining that. Rather it is about OTLP expecting 200, yet Arize Phoenix seems to be returning 204.

So, I think yes both scenarios are about endpoints returning 204 when something else is defined:

• Zipkin /api/v1/spans -> 202
• OTLP /v1/traces -> 200

[mikeldking]

Hey @stfines-clgx ! One of the maintainers of Phoenix. - @codefromthecrypt thanks for the heads up! Commenting to follow along. Feel free to tag me if you find it to be an issue with Phoenix!

[codefromthecrypt]

raised this to follow up on the Phoenix side Arize-ai/phoenix#5963

[mikeldking]

fixed in Arize Phoenix. Thanks for flagging. https://github.com/Arize-ai/phoenix/releases/tag/arize-phoenix-v7.6.0

[codefromthecrypt]

@stfines-clgx I did research into Elastic (due to the earlier note). As far as I can tell, there's no OTLP trace receiver code that can return a 204 (e.g apm server returns 200, and zipkin endpoints aren't present in elastic products either).

So, unless there's anything else, let's get back to this topic about the zipkin exporter.

@slyapustin can you double-check assumptions and mention the product you are exporting traces to? Basically, pick up from this comment.

[slyapustin]

@stfines-clgx

I'm getting mercilessly spammed with 204's hitting the Arize open-telemetry http endpoint when I do local development - DEBUG:urllib3.connectionpool:http://localhost:6006 "POST /v1/traces HTTP/1.1" 204 0 ERROR:opentelemetry.exporter.otlp.proto.http.trace_exporter:Failed to export batch code: 204, reason:

You might need to check which version of the opentelemetry-python is installed in your environment.
status code checks in other parts of the library was fixed in version 1.23.0, so make sure you have version no older than that.