feat: New Garbage Collection Metrics Sampler for .NET 6+

[tippmar-nr]

Description

Adds a new garbage collection metrics sampler that leverages the .NET APIs available in .NET 6 and later, instead of the existing (and somewhat error prone) Event Listener sampler. Available only for applications that target .NET 6 and later.

The new sampler is disabled by default but can be enabled either via newrelic.config:

<appSettings>
    <add key="GCSamplerV2Enabled" value="true" />
</appSettings>

or environment variable:

NEW_RELIC_GC_SAMPLER_V2_ENABLED=1

When the new sampler is enabled, the Event Listener sampler is disabled. Additionally, a supportability metric, Supportability/Dotnet/GCSamplerV2/Enabled is sent when the new sampler is enabled.

The set of garbage collection metrics reported by the new sampler are different from those reported by the Event Listener sampler, as detailed below:

Metric	Event Listener Sampler	Modern Sampler
GC/Handles	✔️
GC/Induced	✔️
GC/Gen0/Size	✔️	✔️
GC/Gen0/Promoted	✔️
GC/Gen0/Collections	✔️	✔️
GC/Gen0/Fragmentation		✔️
GC/Gen1/Size	✔️	✔️
GC/Gen1/Promoted	✔️
GC/Gen1/Collections	✔️	✔️
GC/Gen1/Fragmentation		✔️
GC/Gen2/Size	✔️	✔️
GC/Gen2/Promoted	✔️
GC/Gen2/Survived	✔️
GC/Gen2/Collections	✔️	✔️
GC/Gen2/Fragmentation		✔️
GC/LOH/Size	✔️	✔️
GC/LOH/Survived	✔️
GC/LOH/Collections	✔️	✔️
GC/LOH/Fragmentation		✔️
GC/POH/Size		✔️
GC/POH/Collections	✔️	✔️
GC/POH/Fragmentation		✔️
GC/Heap/Total		✔️
GC/Heap/Committed		✔️
GC/Heap/Allocated		✔️

Author Checklist

• Unit tests, Integration tests, and Unbounded tests completed
• Performance testing completed with satisfactory results (if required)

Reviewer Checklist

• Perform code review
• Pull request was adequately tested (new/existing tests, performance tests)

[chynesNR]

If I'm following the code correctly, the first time through the PreviousSample will be empty, so the first delta report will always be a big spike. Would it be better to skip the first one to get a baseline?

[tippmar-nr]

If I'm following the code correctly, the first time through the PreviousSample will be empty, so the first delta report will always be a big spike. Would it be better to skip the first one to get a baseline?

@chynesNR Originally, I had code in place to wait until at least once GC had completed before sending up any data - which pretty much matches what OTel is doing. I took that out a couple days ago because it didn't seem like our other GC samplers behaved that way. I can go either way with it. @nrcventura

[nrcventura]

If I'm following the code correctly, the first time through the PreviousSample will be empty, so the first delta report will always be a big spike. Would it be better to skip the first one to get a baseline?

@chynesNR Originally, I had code in place to wait until at least once GC had completed before sending up any data - which pretty much matches what OTel is doing. I took that out a couple days ago because it didn't seem like our other GC samplers behaved that way. I can go either way with it. @nrcventura

I think either approach will ultimately be fine. However, if we do wait to report the first value, the charts might be able to scale themselves better.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 92.36641% with 20 lines in your changes missing coverage. Please review.

Project coverage is 81.43%. Comparing base (3afa15f) to head (bc74b10).
Report is 7 commits behind head on main.

Files with missing lines	Patch %	Lines
...Agent/Core/Samplers/GCSamplerV2ReflectionHelper.cs	68.88%	12 Missing and 2 partials ⚠️
....Agent.Extensions/Reflection/VisibilityBypasser.cs	82.60%	4 Missing ⚠️
...wRelic/Agent/Core/Config/BootstrapConfiguration.cs	89.47%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2838      +/-   ##
==========================================
+ Coverage   81.33%   81.43%   +0.09%     
==========================================
  Files         460      464       +4     
  Lines       29239    29496     +257     
  Branches     3231     3261      +30     
==========================================
+ Hits        23781    24019     +238     
- Misses       4666     4682      +16     
- Partials      792      795       +3     

Flag	Coverage Δ
Agent	82.33% <92.36%> (+0.10%)	⬆️
Profiler	73.33% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...elic/Agent/Core/AgentHealth/AgentHealthReporter.cs	87.20% <100.00%> (+0.17%)	⬆️
...c/Agent/Core/Configuration/DefaultConfiguration.cs	89.58% <100.00%> (+0.01%)	⬆️
.../Agent/Core/Configuration/ReportedConfiguration.cs	95.74% <100.00%> (+0.01%)	⬆️
...ic/Agent/Core/DependencyInjection/AgentServices.cs	98.22% <100.00%> (+0.13%)	⬆️
...c/Agent/NewRelic/Agent/Core/Metrics/MetricNames.cs	92.15% <100.00%> (+0.28%)	⬆️
.../Agent/NewRelic/Agent/Core/Samplers/GCSamplerV2.cs	100.00% <100.00%> (ø)
...rc/Agent/NewRelic/Agent/Core/Samplers/GcSampler.cs	80.30% <ø> (-2.03%)	⬇️
.../NewRelic/Agent/Core/Samplers/ImmutableGCSample.cs	100.00% <100.00%> (ø)
...c/Agent/Core/Transformers/GCSampleTransformerV2.cs	100.00% <100.00%> (ø)
...wRelic/Agent/Core/Config/BootstrapConfiguration.cs	98.82% <89.47%> (-1.18%)	⬇️
... and 2 more

... and 2 files with indirect coverage changes