We want to monitor resource usage for all the VMS group by tenants. We do not want to write our own plugin for monitoring, but use ceilometer. So instead of installing the full ceilometer components, we can install just the ceilometer-compute agent on each hypervisor, and then use the pipeline.yaml file to send the collected stats directly to graphite.
Performance issues on saving records to sql and on query.
To create a new graphite publisher, so that ceilometer-compute agent can directly send the metrics to graphite.
publishers: - graphite://127.0.0.1:2003
Working Sample. The metrics are send to graphite as: ceilometer.project_id.resource_id.metric
Working Code: https://github.com/a1git/ceilograph
Code Review: https://review.openstack.org/#/c/103479/2