PYTHON-5846 Consolidate CMAP, heartbeat, and SDAM telemetry into _telemetry.py by blink1073 · Pull Request #2907 · mongodb/mongo-python-driver

blink1073 · 2026-07-01T01:13:05Z

Changes in this PR

Consolidates all APM event publishing and structured logging into pymongo/_telemetry.py, giving OpenTelemetry support a single place to hook into. Adds four new telemetry classes and one helper following the _CommandTelemetry pattern from #2891:

_CmapTelemetry — connection pool and connection lifecycle events, owned by Pool
_HeartbeatTelemetry — server heartbeat events, owned per-check by Monitor
_SdamTelemetry — topology and server description change events, owned by Topology and Server
_ServerSelectionTelemetry — server selection log entries (log-only per spec), constructed per select_server call
log_command_retry — retry log entries for retryable reads and writes

Test Plan

Covered by existing test suites (test_connection_monitoring.py, test_heartbeat_monitoring.py, test_sdam_monitoring_spec.py, test_connection_logging.py, test_server_selection_logging.py, unified format spec tests). No new public API.

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).

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?

…emetry.py Add _CmapTelemetry, _HeartbeatTelemetry, and _SdamTelemetry classes to eliminate the repetitive if-enabled_for_cmap / if-logger.isEnabledFor boilerplate spread across pool.py, monitor.py, topology.py, and server.py.

- Rename _CmapTelemetry._log -> _emit_log for consistency with _CommandTelemetry - Rename _HeartbeatTelemetry.apm_started -> started to match started/succeeded/failed lifecycle - Rename _HeartbeatTelemetry.log_started -> emit_started_log to signal it is the deferred log-only half - Replace is_sdam flag with separate publish/log bool parameters on _CmapTelemetry

_CmapTelemetry now owns connection-creation and checkout durations: connection_created() starts the clock, connection_ready() computes it; checkout_started() starts the clock, checkout_succeeded/failed() compute it. _HeartbeatTelemetry.started() starts the clock; failed() computes its own duration instead of receiving it as a parameter. Removes checkout_started_time from Pool._get_conn, _raise_if_not_ready, and _raise_wait_queue_timeout, and removes AsyncConnection.creation_time.

Use response.awaitable (ground truth from wire) rather than self._awaited (pre-computed before response) for ServerHeartbeatSucceededEvent.awaited. Also fix stale method references in _HeartbeatTelemetry docstring.

…emetry

…seconds)

_CmapTelemetry, _HeartbeatTelemetry, and _SdamTelemetry now compute _should_publish and _should_log as properties that check listener state and logger level at call time rather than caching derived booleans in the constructor.

…listeners

codecov-commenter · 2026-07-01T01:51:24Z

Codecov Report

❌ Patch coverage is 98.92183% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
pymongo/_telemetry.py	98.23%	1 Missing and 3 partials ⚠️

📢 Thoughts on this report? Let us know!

…to _telemetry.py

…y classes; document why long-lived classes use properties

…ion computation in _CmapTelemetry and _HeartbeatTelemetry

…g in _CmapTelemetry

…ng is enabled

NoahStapp · 2026-07-01T17:56:25Z

-                        clientId=self._client._topology_id,
-                        commandName=self._operation,
-                        operationId=self._operation_id,
+                    log_command_retry(


What do you think about a helper method here to avoid duplicating the entire log_command_retry call except for is_write between this and the next usage?

NoahStapp · 2026-07-01T17:56:53Z

 from pymongo.hello import Hello
 from pymongo.lock import _async_create_lock
-from pymongo.logger import _SDAM_LOGGER, _debug_log, _SDAMStatusMessage
+from pymongo.logger import _SDAM_LOGGER, _debug_log


Unused imports.

They were used, I missed one direct usage, now moved to _telemetry.py

NoahStapp · 2026-07-01T18:10:06Z

+        should_publish = self._should_publish
+        # Always record start time: logging or publishing may be enabled by the time
+        # checkout_succeeded or checkout_failed is called to compute the duration.
+        self._checkout_start = time.monotonic()


This is shared across all connections in the pool and is called outside of a lock, introducing a possible race condition that can cause timing inconsistencies.

Updated to keep the timing info in the pool class as it was before.

NoahStapp · 2026-07-01T18:11:00Z

+    def connection_created(self, conn_id: int) -> None:
+        # Always record start time: logging or publishing may be enabled by the time
+        # connection_ready is called to compute the duration.
+        self._conn_created_start = time.monotonic()


Same race condition here as above.

NoahStapp · 2026-07-01T18:24:14Z

-                        clientId=self.description._topology_settings._topology_id,
-                        failure=self._error_message(selector),
-                    )
+                ss.failed(self._error_message(selector))


This will log self.description at the time of creation instead of at the time of failure. If SDAM updates self.description while this loop is waiting and before it calls ss.failed, we log the old description instead of the new.

Co-authored-by: Noah Stapp <noah@noahstapp.com>

- Fix race condition in _CmapTelemetry: remove shared _conn_created_start and _checkout_start slots; connection_ready now takes creation_time from AsyncConnection.creation_time, checkout_started returns the start time and checkout_succeeded/failed accept it as a parameter - Move SRV monitor failure log into _telemetry.py as log_srv_monitor_failure - Add _log_retry helper to _ClientConnectionRetryable to deduplicate log_command_retry call sites - Fix _ServerSelectionTelemetry.failed to accept live topology_description at failure time instead of using the stale snapshot from construction

NoahStapp · 2026-07-01T19:36:53Z

            await self.checkin(conn)

-    def _raise_if_not_ready(self, checkout_started_time: float, emit_event: bool) -> None:
+    def _raise_if_not_ready(self, emit_event: bool, checkout_start: float) -> None:


intentional re-ordering + naming of checkout_started_time?

Restored to original signature

NoahStapp · 2026-07-01T19:37:15Z


    async def _get_conn(
-        self, checkout_started_time: float, handler: Optional[_MongoClientErrorHandler] = None
+        self, handler: Optional[_MongoClientErrorHandler] = None, checkout_start: float = 0.0


Same here. Also, does a default of 0 make sense here?

Restored to original signature

NoahStapp · 2026-07-01T19:37:31Z

-                            self._raise_wait_queue_timeout(checkout_started_time)
-                        self._raise_if_not_ready(checkout_started_time, emit_event=False)
+                            self._raise_wait_queue_timeout(checkout_start)
+                        self._raise_if_not_ready(emit_event=False, checkout_start=checkout_start)


All these uses too.

NoahStapp · 2026-07-01T19:51:07Z

+            **extra,
+        )
+
+    def pool_created(self, non_default_options: dict[str, Any]) -> None:


Need docstrings on these public methods for consistency with the other telemetry classes.

Added :allthedocstrings:

NoahStapp · 2026-07-01T19:55:27Z

-                topologyDescription=self.description,
-                clientId=self.description._topology_settings._topology_id,
-            )
+        ss = _ServerSelectionTelemetry(


Can we pass this through as an argument from select_server so we don't have to make two _ServerSelectionTelemetry per call?

NoahStapp · 2026-07-01T19:56:22Z

@@ -74,6 +69,7 @@ def __init__(
        self._events = None


Can we cut all three of these now that _SdamTelemetry exists?

NoahStapp · 2026-07-01T19:57:10Z

+                remainingTimeMS=remaining_time_ms,
+            )
+
+    def failed(self, failure: str, topology_description: Any = None) -> None:


topology_description appears to always be passed

- Restore original _raise_if_not_ready and _get_conn parameter names/order - Add docstrings to all public methods on _CmapTelemetry and _SdamTelemetry - Remove redundant _publish/_listener/_events attrs from Server.__init__ - Avoid creating two _ServerSelectionTelemetry per select_server call by threading ss through select_servers and _select_server - Make _ServerSelectionTelemetry.failed topology_description a required arg - Move SRV monitor failure log to telemetry.py (log_srv_monitor_failure)

…t_servers/_select_server

blink1073 added 10 commits June 30, 2026 08:27

PYTHON-5846 Fix awaited value in heartbeat succeeded APM event

a30d380

Use response.awaitable (ground truth from wire) rather than self._awaited (pre-computed before response) for ServerHeartbeatSucceededEvent.awaited. Also fix stale method references in _HeartbeatTelemetry docstring.

PYTHON-5846 Add _emit_log helpers to _HeartbeatTelemetry and _SdamTel…

577b594

…emetry

PYTHON-5846 Fix durationMS unit in CMAP log entries (seconds -> milli…

fc143ff

…seconds)

PYTHON-5846 Drop publish param from _HeartbeatTelemetry, derive from …

553dc31

…listeners

PYTHON-5846 Decompose Hello response in _HeartbeatTelemetry.succeeded

0f33a1c

PYTHON-5846 Fix typing: use Hello type in _HeartbeatTelemetry.succeeded

3eefcf7

blink1073 added 6 commits July 1, 2026 06:22

PYTHON-5846 Consolidate server selection logging and retry logging in…

d844557

…to _telemetry.py

PYTHON-5846 Move _should_log check to call site in all telemetry classes

57f4e50

PYTHON-5846 Cache _should_log/_should_publish in short-lived telemetr…

d51f38b

…y classes; document why long-lived classes use properties

PYTHON-5846 Cache _should_log/_should_publish locally and guard durat…

de14677

…ion computation in _CmapTelemetry and _HeartbeatTelemetry

PYTHON-5846 Add comments explaining unconditional start time recordin…

488c750

…g in _CmapTelemetry

PYTHON-5846 Only record heartbeat start time when logging or publishi…

58e966b

…ng is enabled

blink1073 requested a review from NoahStapp July 1, 2026 17:47

blink1073 marked this pull request as ready for review July 1, 2026 17:47

blink1073 requested a review from a team as a code owner July 1, 2026 17:47

NoahStapp requested changes Jul 1, 2026

View reviewed changes

blink1073 and others added 2 commits July 1, 2026 13:32

Update pymongo/asynchronous/mongo_client.py

2746386

Co-authored-by: Noah Stapp <noah@noahstapp.com>

blink1073 requested a review from NoahStapp July 1, 2026 19:06

NoahStapp requested changes Jul 1, 2026

View reviewed changes

blink1073 requested a review from NoahStapp July 1, 2026 20:32

PYTHON-5846 Fix test callers to unpack (servers, ss) tuple from selec…

9fab1db

…t_servers/_select_server

Uh oh!

Conversation

blink1073 commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes in this PR

Test Plan

Checklist

Checklist for Author

Checklist for Reviewer

Uh oh!

codecov-commenter commented Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

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!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

NoahStapp Jul 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

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

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

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!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

blink1073 commented Jul 1, 2026 •

edited

Loading

codecov-commenter commented Jul 1, 2026 •

edited

Loading

NoahStapp Jul 1, 2026 •

edited

Loading