Every 5 minutes, Mira reads the radacct table. If a NAS device has had zero accounting activity for 30 minutes, that's an outage. She opens a critical ticket, computes affected subscribers, and closes it automatically when the device starts talking again.
*/5 * * * * — hits /api/mira/noc-watch. Reads max(accttime) per NAS-IP-Address from radacct.
Configurable per device class. Backhaul routers: 15 min. APs: 30 min. OLTs: 10 min.
Before creating a ticket, checks for an open ticket with [NOC] tag and matching NAS. No spam. No thrash.
Computes count of currently-authenticated subscribers with that NAS-IP in their last session. Shown in the ticket.
When accounting resumes, the ticket closes automatically with downtime duration recorded.
Critical tickets notify #noc-alerts on Slack and page the on-call engineer via email. Configurable per severity.