Hello,

We are trying to collect perf measurements through Ceilometer. The system we have is the following: Openstack Newton Ceilometer client 2.6 Libvirt version 2.< Qemu - KVM Ubuntu 16.04 OS kernel is 4.4.0-96-generic Processor is XEON 2699, which according to intel supports perf measurements.

We have set the " enabled_perf_events= cmt, cpu_cycles, cache_references, cache_misses" in the nova.conf and also the perf measurements in the pipeline.yaml (although we are not sure whether this is required...) as follows: - name: perf_source interval: 240 meters: - "perf.cpu.cycles" - "cpu_l3_cache" - "perf.cache.references" - "perf.cache.misses" - "memory.bandwidth.total" - "memory.bandwidth.local" sinks: - meter_sink

What we get is "Skip pollster perf.cpu.cycles, no new resources found in this cycle" ... Any ideas?

Thank you very much!!