Summary
Internal endpoint the Lambda loader (#7292) calls after each run to report completion metrics against a pre-created metrics row (seeded by the recurring task). Core persists the update and serves aggregates via the telemetry endpoint (#7289).
Depends on: #7276 (connection model)
Consumed by: #7289 (telemetry read)
Called by: #7292 (Lambda loader)
Row seeded by: the recurring task (separate issue)
Endpoint
PATCH /api/v1/organisations/{organisation_pk}/integrations/warehouse-connections/{id}/metrics/{run_id}/
run_id refers to the metrics row pre-created by the recurring task when it invoked the Lambda. The Lambda receives run_id in its event payload and echoes it back here.
Request body
{
"completed_at": "2026-04-20T11:00:12Z",
"events_delivered": 148233,
"batches_dlq": 0,
"error": null
}
completed_at — when the Lambda run finishedevents_delivered — total events successfully COPY'd on this run (0 on failure or empty-prefix ticks)batches_dlq — number of batches that exhausted retries on this run and landed in the DLQerror — short error string if the run itself errored out before completing; null otherwise
Response
204 No Content on success. 404 Not Found if run_id does not exist.
Auth
Service-to-service — not user-authenticated. Mechanism to be decided by implementer (internal API key, IAM/SigV4, or shared secret header).
Storage
Rows in a WarehouseConnectionMetrics table (or similar). Each row represents one loader run and moves through two states:
- Created by the recurring task at invocation time, with
started_at populated and completion fields null
- Updated by this endpoint when the Lambda reports back, populating
completed_at and the metric fields
Rows where started_at is set but completed_at remains null past an expected window indicate an orphaned/crashed Lambda run. #7289 uses this table to compute rolling 24h / 7d aggregates cheaply.
Summary
Internal endpoint the Lambda loader (#7292) calls after each run to report completion metrics against a pre-created metrics row (seeded by the recurring task). Core persists the update and serves aggregates via the telemetry endpoint (#7289).
Depends on: #7276 (connection model)
Consumed by: #7289 (telemetry read)
Called by: #7292 (Lambda loader)
Row seeded by: the recurring task (separate issue)
Endpoint
run_idrefers to the metrics row pre-created by the recurring task when it invoked the Lambda. The Lambda receivesrun_idin its event payload and echoes it back here.Request body
{ "completed_at": "2026-04-20T11:00:12Z", "events_delivered": 148233, "batches_dlq": 0, "error": null }completed_at— when the Lambda run finishedevents_delivered— total events successfullyCOPY'd on this run (0 on failure or empty-prefix ticks)batches_dlq— number of batches that exhausted retries on this run and landed in the DLQerror— short error string if the run itself errored out before completing;nullotherwiseResponse
204 No Contenton success.404 Not Foundifrun_iddoes not exist.Auth
Service-to-service — not user-authenticated. Mechanism to be decided by implementer (internal API key, IAM/SigV4, or shared secret header).
Storage
Rows in a
WarehouseConnectionMetricstable (or similar). Each row represents one loader run and moves through two states:started_atpopulated and completion fields nullcompleted_atand the metric fieldsRows where
started_atis set butcompleted_atremains null past an expected window indicate an orphaned/crashed Lambda run. #7289 uses this table to compute rolling 24h / 7d aggregates cheaply.