Skip to content

VMAFxNoLiveNodes

Meaning. The controller has pending jobs and no registered node for 10 minutes (vmafx_controller_nodes_live is 0 while vmafx_controller_jobs_pending is above 0).

Impact. No submitted job runs. Callers waiting on job results time out.

Diagnose

  1. Are node pods running? kubectl get pods -l app.kubernetes.io/component=node.
  2. Can they reach the controller? A node logs controller client started on success and RegisterNode failed with the reason on failure: kubectl logs <node-pod> | grep -i register. A refusal names the cause: a wrong VMAFX_CONTROLLER_ADDR, a token without the vmafx:node role, or TLS settings that do not match the controller.
  3. Do they stay registered? A node that misses its heartbeats for 60 seconds is evicted; repeated heartbeat failed lines point at the network between node and controller.

Fix

Start or fix the nodes, or correct their controller address, token or TLS settings. Jobs start as soon as one node registers; the alert clears then.

Dashboards: Overview (Live nodes, Pending jobs), Nodes and devices.