- name: d8.cluster-map rules: - alert: ClusterMapHubbleDisconnected expr: max by (pod) (cluster_map_hubble_stream_active == 1 and cluster_map_hubble_connected == 0) for: 15m labels: severity_level: "8" tier: cluster annotations: plk_markup_format: markdown plk_protocol_version: "1" summary: cluster-map has no flows from Hubble Relay. description: | The traffic layer of cluster-map has received no flows for 15 minutes while it is supposed to be collecting. The topology map still works; traffic windows go stale. Check the pod log: `kubectl -n d8-cluster-map logs deploy/cluster-map`, and that hubble-relay is running: `kubectl -n d8-cni-cilium get pods -l app=hubble-relay`. - alert: ClusterMapTrackedTuplesHigh expr: max(cluster_map_traffic_tracked_tuples) > 2000000 for: 30m labels: severity_level: "9" tier: cluster annotations: plk_markup_format: markdown plk_protocol_version: "1" summary: cluster-map keeps more than 2 million connection tuples. description: | Memory of the traffic collector grows with the number of distinct connections in the retention window. Lower `traffic.retentionMinutes` or switch `traffic.mode` to `OnDemand` in the `cluster-map` ModuleConfig.