Telecom · Kubernetes, Zabbix, BSS jobs, NOC tickets
Kubernetes and Zabbix for BSS operations
BSS jobs moved onto Kubernetes. Zabbix watched cluster, node, GPU, and job health. Alerts became tickets in the NOC, not graphs someone happened to have open.
A European operator
The constraint
Health lived in dashboards. A stuck bill job or a node draining overnight waited on whoever was looking. The NOC had a ticket tool. The cluster did not write to it.
What we did
Workloads on Kubernetes. Zabbix on cluster, node, GPU, and the jobs that matter for a bill run. Alerts mapped to the existing ticket path with a queue and an owner, including at 02:00.
How it ran
Cluster, job instrumentation, Zabbix into the ticket tool, runbooks handed to the same NOC. Promon stayed on the queue after go-live.
What was true after
Sev-1s start as tickets. The team that built the map still answers.