RKE2/Rancher: Enable Cilium eBPF syscall metrics (bpf_syscall_duration_seconds)

Observability And Monitoring
#1

Hi everyone,

Thanks in advance for looking into this.

I’m running an RKE2-Rancher cluster in STG using Cilium as the CNI, with Prometheus and Grafana. On the Grafana cilium-metrics dashboard, several panels are empty—specifically the BPF syscall panels.

For example, this query returns no data (or Grafana can’t explore it):

avg(rate(cilium_bpf_syscall_duration_seconds_count{k8s_app=“cilium”, pod=~“cilium.*”}[1m])) by (pod, operation)

All “BPF system call” dashboard panels are blank. I saw in the Cilium documentation that syscall metrics can be disabled by default, so I attempted to enable them, but I still can’t get any data.

Has anyone run into this in an RKE2/Rancher setup and successfully fixed it? If so, could you share what change was needed (HelmChartConfig / values, ServiceMonitor vs annotations, etc.)?

Thanks!

#2

You’ll probably get a quicker response on a Rancher-oriented group. Here’s the Rancher site; it’d be a good place to start looking for that.