feat(instrumentation-mongodb): support aggregation commands and support nested statements

[Fox32]

Which problem is this PR solving?

At Serrala, we have a lot of micro services build with different languages that we observe with OpenTelemetry. We noticed that the mongo instrumentation of Node lacks some details compared to the Java instrumentation:

• Aggregations are not supported (they are just displayed as mongo.command).
• db.statement is sanitized to a degree that it doesn't contain enough details to pin down what the query is doing.

Short description of the changes

This PR adds support for aggregation command. As aggregation pipelines typically have a nested structure to describe the statement, the current way of sanitizing only leaves a low amount of hints what the command is about:

For example:

collection
  .aggregate([
    { $match: { key: 'value' } },
    { $group: { _id: '$a', count: { $sum: 1 } } },
  ])
  .toArray()

Is sanitized to:

{
    aggregate: "?",
    pipeline: "?",
    cursor: "?",
}

The Java instrumentation is handling statement sanitizing differently. Instead of replacing all values of the top level object, it traverses into nested objects and array, keeping the existing structure. This PR implements the same behavior for the Node instrumentation. Therefore db.statement is in this case:

{
    aggregate: "?",
    pipeline: [
      { $match: { key: "?" } },
      { $group: { _id: "?", count: { $sum: "?" } } },
    ],
    cursor: {},
}

While it is possible to configure a customer sanitizer for commands, I think it would better to add support for this upstream for better compatibility to the Java instrumentation.

I fixed some small typos I stumbled upon while reading through the contribution guidelines - I hope it is fine to include them here.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: Fox32 / name: Oliver Sand (ea52e6a, 033f51c)
• ✅ login: pichlermarc / name: Marc Pichler (1d74905, c62e76b)
• ✅ login: JamieDanielson / name: Jamie Danielson (b60d3d3)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 91.43%. Comparing base (dfb2dff) to head (e1a8bed).
Report is 152 commits behind head on main.

❗ Current head e1a8bed differs from pull request most recent head b60d3d3

Please upload reports for the commit b60d3d3 to get more accurate results.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1728      +/-   ##
==========================================
+ Coverage   90.97%   91.43%   +0.45%     
==========================================
  Files         146      139       -7     
  Lines        7492     7179     -313     
  Branches     1502     1449      -53     
==========================================
- Hits         6816     6564     -252     
+ Misses        676      615      -61     

Files	Coverage Δ
...try-instrumentation-mongodb/src/instrumentation.ts	53.95% <100.00%> (+0.68%)	⬆️
...etry-instrumentation-mongodb/src/internal-types.ts	100.00% <100.00%> (ø)

... and 40 files with indirect coverage changes

[Fox32]

@pichlermarc I'm not sure whether the failing tests are related to my changes at all, looks more like a flaky test?

[david-luna]

@Fox32 thanks for contributing to opentelementry :)

there has been recent changes that conflict with your branch (#1760)
resolve the conflicts and push again (ping me if you need help). As a side effect it will run again the CI so we can check if we get the same error in propagation-utils package.

[Fox32]

@david-luna Thanks for notifying me. I updated the PR, can you retrigger the pipeline?

[david-luna]

Sorry I can't trigger the CI

cc: @open-telemetry/javascript-maintainers

[david-luna]

Thanks again @Fox32 for your contribution :)

Semantic conventions define the db.statement as a string value (link). Also the specific section for mongodb does not mention anything about statements (link)

However Java instrumentation and even this one also send an object in some cases (haven't checked other agents). I guess this is something that should be updated in semantic-conventions and we should check if there is an issue for that (or crate one if not).

Saying that this PR looks good :)

[JamieDanielson]

Looks like the merge from main made the change for SemanticAttributes.* changing to SEMATTRS_*, which is what caused the tests to fail. I've just commited those changes are will see how the tests do this time around 😄

[Fox32]

@JamieDanielson thanks for taking care!