Updating Status based on the new spec by eddynaka · Pull Request #1313 · open-telemetry/opentelemetry-dotnet

eddynaka · 2020-09-29T00:26:51Z

Fixes #1311.

Changes

Please provide a brief description of the changes here.

For significant contributions please make sure you have completed the following items:

CHANGELOG.md updated for non-trivial changes
Design discussion issue #
Changes in public API reviewed

eddynaka · 2020-09-29T00:29:45Z

What I did so far:

almost all status is mapped to Error
created a new enum StatusCode with Ok, Error, and Unset.

StatusCanonicalCode is still here, because we need that for grpc

codecov · 2020-09-29T00:47:33Z

Codecov Report

Merging #1313 into master will decrease coverage by 0.22%.
The diff coverage is 92.68%.

@@            Coverage Diff             @@
##           master    #1313      +/-   ##
==========================================
- Coverage   77.88%   77.66%   -0.23%     
==========================================
  Files         226      226              
  Lines        6386     6314      -72     
==========================================
- Hits         4974     4904      -70     
+ Misses       1412     1410       -2

Impacted Files	Coverage Δ
...tation.AspNetCore/Implementation/HttpInListener.cs	`87.50% <ø> (-0.12%)`	⬇️
...ent/Implementation/GrpcClientDiagnosticListener.cs	`83.33% <ø> (-0.46%)`	⬇️
...tp/Implementation/HttpHandlerDiagnosticListener.cs	`78.82% <25.00%> (ø)`
src/OpenTelemetry.Api/Trace/ActivityExtensions.cs	`99.12% <100.00%> (+0.03%)`	⬆️
src/OpenTelemetry.Api/Trace/SpanHelper.cs	`100.00% <100.00%> (+1.38%)`	⬆️
src/OpenTelemetry.Api/Trace/Status.cs	`100.00% <100.00%> (ø)`
...metryProtocol/Implementation/ActivityExtensions.cs	`90.34% <100.00%> (ø)`
...try.Instrumentation.GrpcNetClient/GrpcTagHelper.cs	`100.00% <100.00%> (ø)`
...ient/Implementation/SqlClientDiagnosticListener.cs	`75.36% <100.00%> (ø)`
...mentation/RedisProfilerEntryToActivityConverter.cs	`94.59% <100.00%> (ø)`
... and 3 more

reyang

LGTM once the otlp.status_code is set to int.

eddynaka · 2020-10-10T19:55:27Z

Just adding the label pr:do-not-merge because the spec is changing a little bit. So, waiting the formal changes to occur to finish it. Besides that, everything is done.

commenting OtlpTrace.Status.Type check fixing test undoing change to statuscanonicalcode fixing tests reiley's comments SetStatus now saves enum instead of string fixing changes after conversion fixing redis tests from http spec, if 1xx,2xx,3xx => unset, otherwise error. from rpc spec, if ok => unset, otherwise error. fixing sql tests fixing redis tests

CodeBlanch · 2020-10-12T16:47:54Z

        public static Status ResolveSpanStatusForGrpcStatusCode(int statusCode)
        {
-            var newStatus = Status.Unknown;
+            var status = Status.Error;


I'm not an expert in what gRPC is doing, but this seems off to me. Just took a quick look at the code it seems to set a status code in grpc.status_code, then it sets status from that, and then it removes grpc.status_code. This new version, I feel like we are going to lose some fidelity that was there previously. Maybe we should set status as it is done here, but leave grpc.status_code set to the raw "canonicalcode" value? Maybe we should also move StatusCanonicalCode into gRPC library since it is no longer in line with the spec.

/cc @alanwest

from this part, I'm still waiting a formal change in the spec to see.

I think I didn't change the part where it gets the value from the activity, so it might remain in the same way (the canonicalcode one). And, for the status itself, yeah, that will have only the 3 possibilities from spec.

Still waiting to confirm its behavior...

I guess most likely the gRPC code will be stored in a similar fashion as http.status_code since it has more semantic than the Unset/Ok/Error value.

I would suggest that we merge this PR as-is since the API part looks good, and we can catch the 0.7.0 release train.
The gRPC semantic convention is a lower priority comparing to the API since it is not a blocking issue, and it can be part of the 0.8.0 release or later.

Right now, GRPC already fills the status in the tag as grpc.status_code. So, it won't be a problem at all 👍

Yes, I think we can proceed and merge. Right now, the only part that is not "done" is the rpc part in the spec, since it will change because it was not updated in the first place.

Removing the label and adding to be merged asap

grpc.status_code is added by the library and the removal of the attribute in favor of the spec-approved attribute is intentional. See #1021.

Are we adding back grpc.status_code until the spec change gets sorted out?

I see a suggestion for something like rpc.grpc.status_code here open-telemetry/opentelemetry-specification#1044 (comment).

I agree with @CodeBlanch that it will likely make sense to bring the StatusCanonicalCode into the gRPC instrumentation.

hi @alanwest , it would be strange to remove one tag and add another tag which contains the same value...i would just enable it and try to set that from spec side. for example, so we would need to just enable it in our side.

(just updated the issue to maintain grpc.status_code. With that, we will just have to remove the SetTag => null for that key.

just enabled and i will create a new issue so we can follow the changes in the spec

…/opentelemetry-dotnet into feature/breaking-status

eddynaka requested a review from a team September 29, 2020 00:26