I installed OpenStack AIO with ceilometer and gnocchi using the following commands:
git clone -b 26.1.1 https://opendev.org/openstack/openstack-ansible /opt/openstack-ansible
cd /opt/openstack-ansible
scripts/bootstrap-ansible.sh
export SCENARIO="aio_metal_horizon_aodh_gnocchi_ceilometer"
scripts/bootstrap-aio.sh
cd playbooks
openstack-ansible setup-hosts.yml
openstack-ansible setup-infrastructure.yml
openstack-ansible setup-openstack.yml
After completing the install and starting a test instance, some metrics appeared in "openstack metric list", but the cpu and memory.usage metrics I was most interested in did not. After wasting a lot of time debugging this I finally figured out that openstack-ansible misconfigured ceilometer due to the following lines in the ceilometer.conf.j2 template in the os_ceilometer role:
{% if 'ceilometer_agent_central' in group_names %}
polling_namespaces = central
{% elif 'ceilometer_agent_compute' in group_names %}
polling_namespaces = compute
{% endif %}
In a metal AIO build, group_names contains both ceilometer_agent_central and ceilometer_agent_compute, and the correct config should be:
polling_namespaces = central,compute
but instead I got:
polling_namespaces = central
After fixing the config manually and restarting ceilometer services, metrics started working fine. Or you can add the following to user_variables.yml as a workaround:
ceilometer_ceilometer_conf_overrides:
DEFAULT:
polling_namespaces: central,compute
Fix proposed to branch: master
Review: https:/