Connector cycle telemetry¶
First-class, connector-layer timing and quality statistics for cyclic
connectors. This cluster :satisfies: Connector framework (FEAT_0030).
First-class, connector-layer timing and quality statistics for cyclic
connectors — the intra-bus quantities only the connector can observe.
Per the hybrid measurement split (Hybrid two-layer timing mea... (ADR_0063)), the connector
measures what it alone sees (wire-round duration, cycle-phase wait,
working-counter quality, per-device freshness); the executor separately
measures the cadence of the task that drives the exchange. The two
layers share a |
A cyclic connector shall report, per bus, the duration of the wire
round performed inside The connector holds no clock; the wire-round duration is supplied to it by the bus driver as a pre-computed value (Hybrid two-layer timing mea... (ADR_0063)). A cycle that produced no valid wire round (a hard fault, see Connector push fault semantics (REQ_0267)) contributes no sample. This is distinct from the executor-measured NC-task execute duration,
which brackets |
A cyclic connector shall report, per bus, the duration spent waiting
for the cycle phase inside |
A cyclic connector shall expose a monotonic per-bus counter that
increments on each cycle whose working-counter (or protocol-equivalent
participation check) does not match the expected device set — i.e. the
condition that drives a transition to ConnectorHealth state machine (REQ_0230)’s |
A cyclic connector shall report, per bus, a monotonic count of cycles
that were not all-devices-fresh ( |
Connector cycle statistics shall be available by the same two paths as the executor (Statistics query API (REQ_0103)), exposed on the connector’s own telemetry hook (not folded into the executor’s observation):
Both paths shall be allocation-free on the connector side and shall
not require |
The connector shall emit a push observation (Connector statistics query API (REQ_0265)) once
per Three cycle outcomes shall be distinguishable from the observation:
A |