[blink1073]

PYTHON-6262

Changes in this PR

In load-balanced mode, cursors pin to a single connection and every getMore must run on it. The pinned-connection path bypassed the retryable-read wrapper, so a retryable overload error from the server aborted the cursor instead of being retried.

• Sends pinned getMores through the retryable-read path, retrying on the pinned connection.
• Runs the full test suite in the load-balancer task, in addition to the load-balancer-marked tests.
• Skips the multi-mongos retryable-writes test on load-balanced topologies, where a retry routed to a different mongos is not observable.

Test Plan

Adds an integration test that pins a cursor and fails getMore with a retryable overload error. The cursor drains while the failures stay within the adaptive retry limit and raises once they exceed it. Needs a live server with failCommand enabled.

Checklist

Checklist for Author

• Did you update the changelog (if necessary)?
• Is there test coverage?
• Is any followup work tracked in a JIRA ticket? If so, add link(s). PYTHON-6167 splits the per-suite task families, including separating sync and async tasks.

Checklist for Reviewer

• Does the title of the PR reference a JIRA Ticket?
• Do you fully understand the implementation? (Would you be comfortable explaining how this code works to someone else?)
• Is all relevant documentation (README or docstring) updated?

[Copilot]

🟡 Changes recommended

A retryable overload reply can leave exhaust-stream state stale, causing a receive-only retry to block indefinitely with the default socket timeout.

1 open issue

• Prevent retrying exhaust getMores after the stream ends · New

What changed in this PR

This PR addresses PYTHON-6262 by routing pinned cursor getMore operations through PyMongo’s retryable-read handling, allowing overload retries on the same connection in load-balanced mode.

Changes:

• Passes pinned connection managers through the shared retry machinery.
• Adds regression coverage for pinned retries and skips an incompatible multi-mongos test.
• Expands load-balancer CI coverage and documents the fix.

File	Description
test/​test_retryable_writes.py	Generated sync skip for the multi-mongos test in load-balanced mode.
test/​test_client_backpressure.py	Generated sync coverage for pinned overload retries and retry limits.
test/​asynchronous/​test_retryable_writes.py	Skips the multi-mongos test in load-balanced mode.
test/​asynchronous/​test_client_backpressure.py	Tests pinned overload retries and retry limits.
pymongo/​synchronous/​mongo_client.py	Generated sync integration of pinned connections into retry handling.
pymongo/​mongo_client_shared.py	Stores the pinned connection manager in shared retry state.
pymongo/​asynchronous/​mongo_client.py	Routes pinned reads through retry handling while reusing their connection.
doc/​changelog.rst	Documents the load-balanced getMore retry fix.
.evergreen/​scripts/​test_generate_config.py	Tests the expanded load-balancer marker selection.
.evergreen/​scripts/​run_tests.py	Includes default sync and async suites in load-balancer runs.

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

The new test fails before exercising the retry guard on pre-7.1 mongos/load-balanced deployments and in Stable API require-v1 variants because it lacks version gating and omits the configured server_api option.

2 open issues

• Skip exhaust cursor tests on mongos versions below 7.1 · New
• Pass the configured server API version to the client · New

1 resolved since last review

• Prevent retrying exhaust getMores after the stream ends

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

Pinned-connection retries and receive-only exhaust handling require live-server validation, including a load-balanced topology with failCommand enabled, which could not be exercised in this review environment.

No open issues 🎉

2 resolved since last review

• Pass the configured server API version to the client
• Skip exhaust cursor tests on mongos versions below 7.1

_{Review effort: Balanced}