Ceilometer

The ceilometer project aims to deliver a unique point of contact for billing systems to acquire all of the measurements they need to establish customer billing, across all current OpenStack core components with work underway to support future OpenStack components.

Sample pillars

Ceilometer API/controller node

ceilometer:
  server:
    enabled: true
    version: havana
    cluster: true
    secret: pwd
    bind:
      host: 127.0.0.1
      port: 8777
    identity:
      engine: keystone
      host: 127.0.0.1
      port: 35357
      tenant: service
      user: ceilometer
      password: pwd
    message_queue:
      engine: rabbitmq
      host: 127.0.0.1
      port: 5672
      user: openstack
      password: pwd
      virtual_host: '/openstack'
      rabbit_ha_queues: true
    database:
      engine: mongodb
      host: 127.0.0.1
      port: 27017
      name: ceilometer
      user: ceilometer
      password: pwd

Ceilometer Graphite publisher

ceilometer:
  server:
    enabled: true
    publisher:
      graphite:
        enabled: true
        host: 10.0.0.1
        port: 2003

Ceilometer compute agent

ceilometer:
  agent:
    enabled: true
    version: havana
    secret: pwd
    identity:
      engine: keystone
      host: 127.0.0.1
      port: 35357
      tenant: service
      user: ceilometer
      password: pwd
    message_queue:
      engine: rabbitmq
      host: 127.0.0.1
      port: 5672
      user: openstack
      password: pwd
      virtual_host: '/openstack'
      rabbit_ha_queues: true

Read more

• https://wiki.openstack.org/wiki/Ceilometer
• http://docs.openstack.org/developer/ceilometer/install/manual.html
• http://docs.openstack.org/developer/ceilometer/
• https://fedoraproject.org/wiki/QA:Testcase_OpenStack_ceilometer_install
• https://github.com/spilgames/ceilometer_graphite_publisher
• http://engineering.spilgames.com/using-ceilometer-graphite/

Things to improve/consider

• Graphite publisher http://engineering.spilgames.com/using-ceilometer-graphite/
• Juno additions - Split Events/Meters and Alarms databases, Polling angets are HA now, active/Activr Workload partitioning to central agents
• Kilo additions - Splint Events - Meters - Agents, notification agents are HA now (everything is HA now), events - elastic search
• User notifier publisher vs rpc publisher (Juno+)
• Enable jittering (rendom delay) to polling. (Kilo+)
• Collect what you need - pipeline.yaml, tweak polling interval (Icehouse+)
• add more agents as load inceases (Juno+)
• Avoid open-ended queries - query on a time range
• Install api behind mod_wsgi, tweak wsgi daemon - threads and processes
• Set TTL - expire data to minimise database size
• Run Mongodb on separate node - use sharding and replica-sets

Deployment scenarios

• Lambda design - use short term and long term databases in the same time
• Data segragation - separatem
• JSON files - Apache spark
• Fraud detection - proprietary alarming system
• Custom consumers - kafka - Apache Storm (kilo+)
• Debugging - Collecttions - Elastic serach - Kibana
• Noisy services - Multiple notification buses