Geneva exporter: Support resource attributes in log exporter by mattsains · Pull Request #3646 · open-telemetry/opentelemetry-dotnet-contrib

mattsains · 2025-12-30T21:49:07Z

Related to #3214

Changes

Allow specifying resource attributes to export in Geneva Log exporter.
Changed behaviour of trace exporter to make resource attributes and prepopulated fields mutually exclusive

Merge requirement checklist

CONTRIBUTING guidelines followed (license requirements, 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)

linux-foundation-easycla · 2025-12-30T21:49:15Z

The committers listed above are authorized under a signed CLA.

✅ login: mattsains / name: Matthew Sainsbury (0393845, 0538e92, 1469019, 1b8bec5, 2d6fc06, 4fe80cb, 50b68dd, 5805669, 604fd2b, 6149cb1, 630b3ab, 6c2a3eb, 70d310c, 988a757, 98a159f, 9987054, a53b3bf, ccfb2fd, d6c3940, e369add)
✅ login: reyang / name: Reiley Yang (4fe80cb)

codecov · 2025-12-30T21:50:55Z

Codecov Report

❌ Patch coverage is 92.70833% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.72%. Comparing base (8ca7383) to head (1469019).
⚠️ Report is 23 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...rter.Geneva/Internal/MsgPack/MsgPackLogExporter.cs	88.88%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3646      +/-   ##
==========================================
+ Coverage   71.52%   71.72%   +0.19%     
==========================================
  Files         455      445      -10     
  Lines       17617    17624       +7     
==========================================
+ Hits        12601    12640      +39     
+ Misses       5016     4984      -32

Flag	Coverage Δ
unittests-Exporter.Geneva	`54.15% <92.70%> (+0.37%)`	⬆️
unittests-Instrumentation.Cassandra	`?`

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

Files with missing lines	Coverage Δ
...Telemetry.Exporter.Geneva/GenevaExporterOptions.cs	`89.18% <ø> (ø)`
...OpenTelemetry.Exporter.Geneva/GenevaLogExporter.cs	`90.38% <100.00%> (+2.88%)`	⬆️
...enTelemetry.Exporter.Geneva/GenevaTraceExporter.cs	`86.04% <100.00%> (+3.99%)`	⬆️
...er.Geneva/Internal/MsgPack/MsgPackTraceExporter.cs	`94.21% <100.00%> (+1.40%)`	⬆️
...rter.Geneva/Internal/MsgPack/MsgPackLogExporter.cs	`92.52% <88.88%> (+0.15%)`	⬆️

... and 13 files with indirect coverage changes

🚀 New features to boost your workflow:

📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

mattsains

A while ago we decided that the resource attributes service.name and service.instanceId would be mapped to the message pack data even if they are not present in ResourceFieldNames. I was just thinking and realised that if someone had set cloud.role or cloud.roleInstance in PrepopulatedFields, then after my change, they will not have their logs exported, because the prepopulated field will conflict with the "built-in" resource attribute mapping for service.name and service.instanceId (which gets mapped to cloud.role and cloud.roleInstance even if you didn't specify it in ResourceFieldNames)

what do you think I should do? I can check to see if it's in resource attributes before I accept it as a prepopulated field, and then I guess just drop the prepopulated field?

mattsains · 2025-12-31T17:58:05Z

A while ago we decided that the resource attributes service.name and service.instanceId would be mapped to the message pack data even if they are not present in ResourceFieldNames. I was just thinking and realised that if someone had set cloud.role or cloud.roleInstance in PrepopulatedFields, then after my change, they will not have their logs exported, because the prepopulated field will conflict with the "built-in" resource attribute mapping for service.name and service.instanceId (which gets mapped to cloud.role and cloud.roleInstance even if you didn't specify it in ResourceFieldNames)

what do you think I should do? I can check to see if it's in resource attributes before I accept it as a prepopulated field, and then I guess just drop the prepopulated field?

I have updated the code and the tests to replace a prepopulated field with the equivalent auto-mapped resource attribute (ie., service.name, service.instanceId). I also added a test for this behaviour

rajkumar-rangaraj · 2025-12-31T22:22:15Z

src/OpenTelemetry.Exporter.Geneva/Internal/MsgPack/MsgPackLogExporter.cs

+        if (this.prepopulatedFields != null)
        {
-            for (var i = 0; i < this.prepopulatedFieldKeys.Count; i++)
+            foreach (var field in this.prepopulatedFields)


Existing implementation uses for loop to avoid the allocation from foreach.

as discussed, this isn't a concern because this foreach is unlikely to have a heap allocation

If I recall correctly, foreach inherently cause heap allocation.

It's a major goal to keep heap allocation as low as possible for Geneva exporter. Can you run a benchmark test in Benchmark folder to confirm there is no heap allocation introduced by foreach? (PR #1504 also included a mini benchmark code)

I can't find any benchmarking code in that PR but if you can point me to a specific benchmark test I'd be happy to run it and post the results here

There are two places you can reference:

opentelemetry-dotnet-contrib/test/OpenTelemetry.Exporter.Geneva.Benchmarks/README.md

Line 3 in 593fb31

Navigate to `./test/OpenTelemetry.Exporter.Geneva.Benchmarks` directory and run

Ctrl+F "Log Benchmarks Code" on page [Exporter.Geneva] Perf improvement on .NET 8 #1504 and click on it

Oops, I don't know how I missed that collapseable section in the PR.

Here's the results of the benchmark on my branch:

Method IncludeFormattedMessage Mean Error StdDev Gen0 Allocated

LoggerWithMessageTemplate False 944.3 ns 8.32 ns 7.78 ns - 104 B

LoggerWithDirectLoggerAPI False 1,136.2 ns 10.99 ns 8.58 ns 0.0057 320 B

LoggerWithSourceGenerator False 933.9 ns 11.04 ns 10.32 ns - 64 B

SerializeLogRecord False 316.7 ns 1.98 ns 1.76 ns - -

Export False 638.5 ns 9.73 ns 7.59 ns - -

LoggerWithMessageTemplate True 961.9 ns 8.06 ns 7.54 ns 0.0010 104 B

LoggerWithDirectLoggerAPI True 1,117.1 ns 6.50 ns 5.76 ns 0.0057 320 B

LoggerWithSourceGenerator True 929.5 ns 8.82 ns 8.25 ns - 64 B

SerializeLogRecord True 315.9 ns 0.87 ns 0.73 ns - -

Export True 641.1 ns 7.51 ns 5.86 ns - -

I also ran the tests on main branch. This confirmed that the PR didn't increase memory allocation. Interestingly, the memory allocation for .NET 10 is higher than .NET 8 (both on main branch):

.NET 8

Method IncludeFormattedMessage Mean Error StdDev Gen0 Allocated

LoggerWithMessageTemplate False 1,142.1 ns 22.74 ns 42.16 ns 0.0038 104 B

LoggerWithDirectLoggerAPI False 1,331.4 ns 26.41 ns 36.15 ns 0.0114 320 B

LoggerWithSourceGenerator False 1,122.7 ns 17.91 ns 28.41 ns 0.0019 64 B

SerializeLogRecord False 434.3 ns 8.42 ns 8.27 ns - -

Export False 773.8 ns 15.53 ns 35.04 ns - -

LoggerWithMessageTemplate True 1,132.5 ns 22.65 ns 44.71 ns 0.0038 104 B

LoggerWithDirectLoggerAPI True 1,285.8 ns 24.61 ns 23.02 ns 0.0114 320 B

LoggerWithSourceGenerator True 1,129.4 ns 22.21 ns 24.69 ns 0.0019 64 B

SerializeLogRecord True 434.2 ns 5.31 ns 4.71 ns - -

Export True 769.2 ns 15.25 ns 31.84 ns - -

.NET 10

Method IncludeFormattedMessage Mean Error StdDev Median Gen0 Allocated

LoggerWithMessageTemplate False 1,042.3 ns 20.88 ns 60.58 ns 1,045.5 ns 0.0038 112 B

LoggerWithDirectLoggerAPI False 1,242.6 ns 24.58 ns 47.36 ns 1,222.2 ns 0.0114 320 B

LoggerWithSourceGenerator False 1,063.2 ns 21.24 ns 44.33 ns 1,051.1 ns 0.0019 64 B

SerializeLogRecord False 383.3 ns 7.46 ns 9.70 ns 379.8 ns - -

Export False 745.4 ns 14.88 ns 32.35 ns 741.4 ns - -

LoggerWithMessageTemplate True 1,102.1 ns 31.75 ns 92.60 ns 1,082.8 ns 0.0038 112 B

LoggerWithDirectLoggerAPI True 1,219.1 ns 24.17 ns 33.08 ns 1,213.6 ns 0.0114 320 B

LoggerWithSourceGenerator True 1,069.4 ns 23.99 ns 69.61 ns 1,065.7 ns 0.0019 64 B

SerializeLogRecord True 369.3 ns 4.97 ns 4.41 ns 368.6 ns - -

Export True 763.3 ns 15.09 ns 41.57 ns 755.1 ns - -

Interesting, could you look into where the delta is coming from in .NET 10 comparing to 8?

The 8 bytes come from the size increase of Microsoft.Extensions.Logging.FormattedLogValues, which is created via new in LoggerExtensions.LogInformation.

I checked the source code: in .NET 10, an additional field _cachedToString of type string? was introduced in FormattedLogValues: .NET 10 code vs .NET 8 code. This field is a reference, so it's 8 bytes on 64-bit CLR.

mattsains · 2026-01-02T17:14:01Z

We talked more about the prepopulated/resource attribute thing, and decided that the mapping can only happen when prepopulated fields have not been specified. Otherwise, the default resource will silently overwrite prepopulated values.

src/OpenTelemetry.Exporter.Geneva/Internal/MsgPack/MsgPackLogExporter.cs

xiang17 · 2026-01-05T21:43:00Z

src/OpenTelemetry.Exporter.Geneva/Internal/MsgPack/MsgPackLogExporter.cs

+        if (this.prepopulatedFields != null)
        {
-            for (var i = 0; i < this.prepopulatedFieldKeys.Count; i++)
+            foreach (var field in this.prepopulatedFields)


If I recall correctly, foreach inherently cause heap allocation.

It's a major goal to keep heap allocation as low as possible for Geneva exporter. Can you run a benchmark test in Benchmark folder to confirm there is no heap allocation introduced by foreach? (PR #1504 also included a mini benchmark code)

src/OpenTelemetry.Exporter.Geneva/CHANGELOG.md

Co-authored-by: Reiley Yang <reyang@microsoft.com>

rajkumar-rangaraj · 2026-01-12T17:40:42Z

src/OpenTelemetry.Exporter.Geneva/GenevaLogExporter.cs

        if (useMsgPackExporter)
        {
-            var msgPackLogExporter = new MsgPackLogExporter(options);
+            var msgPackLogExporter = new MsgPackLogExporter(options, () =>


Should this have a null check? If ParentProvider hasn't been set when export is called (possible during SDK initialization edge cases), this will throw NRE. Consider:

() => this.ParentProvider?.GetResource() ?? Resource.Empty

I'll change it but I notice that if ParentProvider can really be null, then the existing metric code is also broken: https://github.com/open-telemetry/opentelemetry-dotnet-contrib/blob/main/src/OpenTelemetry.Exporter.Geneva/Metrics/GenevaMetricExporter.cs#L112

What does this mean? Is it a bug or not?

This is not a bug; the wording may have caused the confusion.
In the existing implementation, this.ParentProvider.GetResource() itself already returns Resource.Empty when no provider is set, so it is safe. We don’t need to add an explicit null check here or manually set Resource.Empty again.

Thanks @rajkumar-rangaraj! Then this sounds wrong and should be reverted back.

I'll change it

Yes, this needs a fix. I will send a follow-up PR.

github-actions bot added comp:exporter.geneva Things related to OpenTelemetry.Exporter.Geneva perf Performance related labels Dec 30, 2025

mattsains added 6 commits December 30, 2025 13:49

implement resource attributes for geneva logs

e369add

fix indentation

d6c3940

add tests for cloud extension

1b8bec5

a bunch of changes to re-up the branch

604fd2b

more tests

70d310c

fixed all tests

5805669

mattsains force-pushed the logs-resources branch from 86ba956 to 44fd0df Compare December 30, 2025 21:49

mattsains added 4 commits December 30, 2025 13:51

fix test

a53b3bf

proof reading

6149cb1

.net format

988a757

changelog

ccfb2fd

mattsains force-pushed the logs-resources branch from 44fd0df to ccfb2fd Compare December 30, 2025 21:51

mattsains added 3 commits December 30, 2025 13:59

fix test

0538e92

fix test

9987054

fix linux test

2d6fc06

mattsains commented Dec 30, 2025

View reviewed changes

mattsains marked this pull request as ready for review December 30, 2025 22:42

mattsains requested a review from a team as a code owner December 30, 2025 22:42

add test for auto-mapped resource attrs

98a159f

github-actions bot assigned rajkumar-rangaraj and xiang17 Dec 31, 2025

github-actions bot requested review from rajkumar-rangaraj and xiang17 December 31, 2025 17:57

fix tests

50b68dd

rajkumar-rangaraj reviewed Dec 31, 2025

View reviewed changes

gate cloud extension mapping by prepopulated fields

630b3ab

xiang17 reviewed Jan 5, 2026

View reviewed changes

reyang reviewed Jan 7, 2026

View reviewed changes

src/OpenTelemetry.Exporter.Geneva/CHANGELOG.md Outdated Show resolved Hide resolved

mattsains and others added 3 commits January 7, 2026 11:41

Update src/OpenTelemetry.Exporter.Geneva/CHANGELOG.md

4fe80cb

Co-authored-by: Reiley Yang <reyang@microsoft.com>

keep using frozen dictionary as much as practical

0393845

fix semicolon

6c2a3eb

xiang17 approved these changes Jan 9, 2026

View reviewed changes

rajkumar-rangaraj reviewed Jan 12, 2026

View reviewed changes

respond to PR comments

1469019

mattsains force-pushed the logs-resources branch from 3d6037a to 1469019 Compare January 13, 2026 18:10

rajkumar-rangaraj approved these changes Jan 13, 2026

View reviewed changes

rajkumar-rangaraj added this pull request to the merge queue Jan 13, 2026

Merged via the queue into open-telemetry:main with commit 197c092 Jan 13, 2026
78 checks passed

This was referenced Jan 21, 2026

[release] Prepare release Exporter.Geneva-1.15.0 #3737

Merged

Remove redundant Resource.Empty fallback in GetResource() calls #3806

Merged

Method	IncludeFormattedMessage	Mean	Error	StdDev	Gen0	Allocated
LoggerWithMessageTemplate	False	944.3 ns	8.32 ns	7.78 ns	-	104 B
LoggerWithDirectLoggerAPI	False	1,136.2 ns	10.99 ns	8.58 ns	0.0057	320 B
LoggerWithSourceGenerator	False	933.9 ns	11.04 ns	10.32 ns	-	64 B
SerializeLogRecord	False	316.7 ns	1.98 ns	1.76 ns	-	-
Export	False	638.5 ns	9.73 ns	7.59 ns	-	-
LoggerWithMessageTemplate	True	961.9 ns	8.06 ns	7.54 ns	0.0010	104 B
LoggerWithDirectLoggerAPI	True	1,117.1 ns	6.50 ns	5.76 ns	0.0057	320 B
LoggerWithSourceGenerator	True	929.5 ns	8.82 ns	8.25 ns	-	64 B
SerializeLogRecord	True	315.9 ns	0.87 ns	0.73 ns	-	-
Export	True	641.1 ns	7.51 ns	5.86 ns	-	-

Method	IncludeFormattedMessage	Mean	Error	StdDev	Gen0	Allocated
LoggerWithMessageTemplate	False	1,142.1 ns	22.74 ns	42.16 ns	0.0038	104 B
LoggerWithDirectLoggerAPI	False	1,331.4 ns	26.41 ns	36.15 ns	0.0114	320 B
LoggerWithSourceGenerator	False	1,122.7 ns	17.91 ns	28.41 ns	0.0019	64 B
SerializeLogRecord	False	434.3 ns	8.42 ns	8.27 ns	-	-
Export	False	773.8 ns	15.53 ns	35.04 ns	-	-
LoggerWithMessageTemplate	True	1,132.5 ns	22.65 ns	44.71 ns	0.0038	104 B
LoggerWithDirectLoggerAPI	True	1,285.8 ns	24.61 ns	23.02 ns	0.0114	320 B
LoggerWithSourceGenerator	True	1,129.4 ns	22.21 ns	24.69 ns	0.0019	64 B
SerializeLogRecord	True	434.2 ns	5.31 ns	4.71 ns	-	-
Export	True	769.2 ns	15.25 ns	31.84 ns	-	-

Method	IncludeFormattedMessage	Mean	Error	StdDev	Median	Gen0	Allocated
LoggerWithMessageTemplate	False	1,042.3 ns	20.88 ns	60.58 ns	1,045.5 ns	0.0038	112 B
LoggerWithDirectLoggerAPI	False	1,242.6 ns	24.58 ns	47.36 ns	1,222.2 ns	0.0114	320 B
LoggerWithSourceGenerator	False	1,063.2 ns	21.24 ns	44.33 ns	1,051.1 ns	0.0019	64 B
SerializeLogRecord	False	383.3 ns	7.46 ns	9.70 ns	379.8 ns	-	-
Export	False	745.4 ns	14.88 ns	32.35 ns	741.4 ns	-	-
LoggerWithMessageTemplate	True	1,102.1 ns	31.75 ns	92.60 ns	1,082.8 ns	0.0038	112 B
LoggerWithDirectLoggerAPI	True	1,219.1 ns	24.17 ns	33.08 ns	1,213.6 ns	0.0114	320 B
LoggerWithSourceGenerator	True	1,069.4 ns	23.99 ns	69.61 ns	1,065.7 ns	0.0019	64 B
SerializeLogRecord	True	369.3 ns	4.97 ns	4.41 ns	368.6 ns	-	-
Export	True	763.3 ns	15.09 ns	41.57 ns	755.1 ns	-	-

Conversation

mattsains commented Dec 30, 2025

Changes

Merge requirement checklist

Uh oh!

linux-foundation-easycla bot commented Dec 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Dec 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

mattsains left a comment

Choose a reason for hiding this comment

Uh oh!

mattsains commented Dec 31, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

.NET 8

.NET 10

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mattsains commented Jan 2, 2026

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

linux-foundation-easycla bot commented Dec 30, 2025 •

edited

Loading

codecov bot commented Dec 30, 2025 •

edited

Loading