Cron log lines reach Datadog, and the health check watches the Accordion feed
Shipped 2026-10-26
Job logs were never collected
Every scheduled job runs as a child process. The trigger buffered its output
and kept only the CRON_SUMMARY line, so nothing a job logged reached Datadog:
in the 7 days to 2026-10-26 the trigger’s own cron_finished appeared 400
times and the health job’s hourly health_check line not once.
The trigger now forwards each job’s logger lines when the job exits, adding a
cron field with the job’s name. They keep their original timestamp. Progress
text and CRON_SUMMARY are not forwarded.
Health check signals
- Each failed check logs
health.check_failedat error, withcheck,operation,messageandgroup_id. This is what the Datadog monitor will watch. - A failure of the whole run is
health.run_failed; a Slack post that fails ishealth.summary_post_failed(both were unprefixed before). - The health job still posts failures to Slack. That stops once the Datadog monitors are live, with the other notifiers.
Accordion feed check
New check accordionFeed: it fails when no Accordion booking has been stored
for 26 hours. It reads stored bookings rather than connector_history, because
a request rejected at the signature check still writes a history row.
Last updated on