feat: urllib3 instrumentation

[jy-tan]

Summary

Adds instrumentation for urllib3, the low-level HTTP client library that powers requests and other HTTP libraries.

Changes

New instrumentation (drift/instrumentation/urllib3/)

• Patches PoolManager.urlopen() and HTTPConnectionPool.urlopen() to capture HTTP requests
• Supports both RECORD and REPLAY modes with mock response generation

Auto-initialization

• Automatically initializes when urllib3 is imported (via _init_auto_instrumentations)

E2E test suite

• Docker-based test infrastructure with 19 test cases covering GET, POST, PUT, PATCH, DELETE, chained requests, timeouts, retries, redirects, and more

Notable Implementation Details

• Streaming response handling: When preload_content=False is used, response body capture is skipped to avoid consuming the stream before the application reads it. A warning is logged to inform users that replay may return an empty body for these requests. See drift/instrumentation/urllib3/notes.md for more details.
• Double-span prevention: Checks for active spans from RequestsInstrumentation or HttpxInstrumentation to avoid creating duplicate client spans when urllib3 is called internally by those libraries.

[cubic-dev-ai]

1 issue found across 13 files

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="drift/instrumentation/urllib3/e2e-tests/src/app.py">

<violation number="1" location="drift/instrumentation/urllib3/e2e-tests/src/app.py:228">
P2: Close the HTTPSConnectionPool in a finally block so the connection is released even when the request or JSON decoding raises; otherwise failed calls leak sockets.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}