fix(llmobs): reuse shared conn

[Yun-Kim]

Resolves #13336. Credit to @IAL32 and a cherry-pick from #13338.

When we made the jump from using the shared HTTPWriter to our own BaseLLMObsWriter class to submit spans and evals #12966, we used our own _get_connection() to return HTTP/HTTPS connections. However we forgot to include UDSHTTP connection (for the unix socket case), which means we broke UDS support until now.

Why was this a problem in the first place?

We used our own _get_connection() in #12966 because of an issue where creating the shared HTTPConnection helper class was leading to MRO superclass constructor issues in our tests. At the time we thought this was due to the shared HTTPConnection helper class having multiple superclasses and an issue with Python 3.10 in general, but this turns out to be due to vcrpy mocking HTTPConnection entirely and only being an issue in tests that rely on vcrpy. This PR makes some changes to avoid using vcrpy when not necessary, and making better assertions to ensure that spans are being sent (not necessary in most tests to have them be accepted).

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

releasenotes/notes/fix-llmobs-unix-agent-2b959d436e6474c3.yaml          @DataDog/apm-python
ddtrace/llmobs/_writer.py                                               @DataDog/ml-observability
tests/llmobs/test_llmobs_eval_metric_agent_writer.py                    @DataDog/ml-observability
tests/llmobs/test_llmobs_eval_metric_agentless_writer.py                @DataDog/ml-observability
tests/llmobs/test_llmobs_evaluator_runner.py                            @DataDog/ml-observability
tests/llmobs/test_llmobs_span_agent_writer.py                           @DataDog/ml-observability
tests/llmobs/test_llmobs_span_agentless_writer.py                       @DataDog/ml-observability

[github-actions]

Bootstrap import analysis

Comparison of import times between this PR and base.

Summary

The average import time from this PR is: 234 ± 3 ms.

The average import time from base is: 235 ± 2 ms.

The import time difference between this PR and base is: -0.8 ± 0.1 ms.

Import time breakdown

The following import paths have shrunk:

ddtrace.auto 1.844 ms (0.79%)

ddtrace.bootstrap.sitecustomize 1.174 ms (0.50%)

ddtrace.bootstrap.preload 1.174 ms (0.50%)

ddtrace.internal.products 1.174 ms (0.50%)

ddtrace.internal.remoteconfig.client 0.614 ms (0.26%)

ddtrace 0.670 ms (0.29%)

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-05-06 23:31:11

Comparing candidate commit 3fdca66 in PR branch yunkim/llmobs-writer-fix-conn with baseline commit ea27de2 in branch main.

Found 0 performance improvements and 2 performance regressions! Performance is the same for 506 metrics, 4 unstable metrics.

scenario:iast_aspects-format_map_aspect

• 🟥 execution_time [+503.240ns; +567.383ns] or [+15.293%; +17.242%]

scenario:iast_aspects-ospathsplitdrive_aspect

• 🟥 execution_time [+400.478ns; +539.554ns] or [+11.070%; +14.914%]

[Kyle-Verhoog]

just a thought about the change in logs and the implication it might have. It'd also be good to have a better validation than logs for successful transmission but that's a bit larger of an ask outside the scope here

[github-actions]

The backport to 2.21 failed:

The process '/usr/bin/git' failed with exit code 1

To backport manually, run these commands in your terminal:

# Fetch latest updates from GitHub
git fetch
# Create a new working tree
git worktree add .worktrees/backport-2.21 2.21
# Navigate to the new working tree
cd .worktrees/backport-2.21
# Create a new branch
git switch --create backport-13339-to-2.21
# Cherry-pick the merged commit of this pull request and resolve the conflicts
git cherry-pick -x --mainline 1 479699d87a4794d808d854435f3df6ba0e83fb33
# Push it to GitHub
git push --set-upstream origin backport-13339-to-2.21
# Go back to the original working tree
cd ../..
# Delete the working tree
git worktree remove .worktrees/backport-2.21

Then, create a pull request where the base branch is 2.21 and the compare/head branch is backport-13339-to-2.21.