Following investigation into DRIVERS-2677, it was discovered that the Node driver measures heartbeat duration incorrectly. On initial creation of a monitoring connection, the heartbeat duration calculation includes the time taken to establish the initial TCP connection as well as the time to send a hello/legacy hello and receive a response from the server. This goes against the definition of the heartbeat duration for ServerHeartbeatSucceeded and ServerHeartbeatFailed events as laid out in the SDAM Events API spec.

From local testing, when connecting to an Atlas cluster with the driver, this consistently results in initial heartbeat durations of ~200ms with median durations of 21ms.

Use Case

As a Driver Engineer
I want to measure heartbeat duration from just before the call to hello/legacy hello
So that heartbeat duration is measured accurately.

User Impact

• None

Dependencies

• upstream and/or downstream requirements and timelines to bear in mind

Unknowns

• Will this clash with CSOT implementation?

Acceptance Criteria

Implementation Requirements

• Update checkServer in src/sdam/monitor.ts to properly record duration of heartbeat as the time between sending a hello and receiving a response in all cases

Testing Requirements

• Write unit tests that assert that the first heartbeat has a duration that is more consistent with subsequent heartbeats.
  • Using mock local server so that it is integration-like

Documentation Requirements

• None

Follow Up Requirements

• additional tickets to file, required releases, etc