bump grpc client version to fix unobserved exception

[masonyc]

Fixes #4570

Changes

Bump grpc net client to 2.53

Merge requirement checklist

• CONTRIBUTING guidelines followed (nullable enabled, static analysis, etc.)
• Unit tests added/updated
• Appropriate CHANGELOG.md files updated for non-trivial changes
• Changes in public API reviewed (if applicable)

[codecov]

Codecov Report

Merging #4573 (7d9cd8a) into main (bd0d8d6) will increase coverage by 0.00%.
The diff coverage is n/a.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4573   +/-   ##
=======================================
  Coverage   84.86%   84.87%           
=======================================
  Files         313      313           
  Lines       12604    12604           
=======================================
+ Hits        10697    10698    +1     
+ Misses       1907     1906    -1     

see 2 files with indirect coverage changes

[vishweshbankwar]

@masonyc - Could you please also update the changelog for otlp exporter https://github.com/open-telemetry/opentelemetry-dotnet/blob/main/src/OpenTelemetry.Exporter.OpenTelemetryProtocol/CHANGELOG.md

[Kielek]

Lets wait a bit, I do not have a chance to verify impact on .NET AutoInstrumentation. I suppose it will be significant.

[Kielek]

Finally, I have find the time to review impact.

For Auto Instrumentation, it will leads to reduce set of application which can be auto-instrumented without adjustment for dependencies. Similar discussion was done for #4407

Especially, we will have to increase minimal instrumented version Grpc.Client.Net from 2.43.0 to the upgraded one.

As 2.45 contains obvious bug-fix, I see the reason to do this, but I have doubts if it should go up to 2.53.0.

@alanwest, you were mainly checking what we can do for protobuf. I suppose that you have also good insight for this particulat package.

EDIT: Similar thing for #4576 (if the user is using .NET version which brings 7.0.1 or 7.0.0).

[alanwest]

@masonyc thanks for your patience! We spoke about this PR today in our weekly meeting.

Regarding @Kielek's comment:

As 2.45 contains obvious bug-fix, I see the reason to do this, but I have doubts if it should go up to 2.53.0.

We agreed a more conservative bump to 2.45 is preferable, but before we do that, I wanted to understand the issue a bit more.

I haven't studied this issue closely enough to understand in what circumstance this bug manifests itself with OpenTelemetry .NET. My thought is that if we understand the bug a bit more then maybe we could find a way to fix it by changing how we use gRPC client to get around the bug rather than bumping the package version.

I've briefly reviewed grpc/grpc-dotnet#1656, but the repro it provides does not resemble how we currently invoke gRPC client, so it left us wondering how this bug is manifested in OpenTelemetry .NET.

@masonyc do you have a repro you could share generating this bug using OpenTelemetry .NET?

[rwkarg]

It isn’t specific to how the gRPC client is being used. It’s a Task inside the internal client code that isn’t explicitly observed when there is an exception making a call (such as a transient network issue). If you’re making a call, then the internal code path that triggers the gRCP client behavior mentioned in the issue will be exercised.

[alanwest]

It’s a Task inside the internal client code

Gotcha. I've had a chance to look more closely and this makes sense now.

@masonyc please change to 2.45 and I think we're good to merge.