ha-guide-draft intro-ha-key-concepts.html Queue Workers

[Shannon Mitchell]

One thing that I see missing from the key concepts that isn't really covered is the concept of queue workers. These are services don't listen for incoming connections, so using a load balancing or resource cluster management solution might not really apply. These applications connect to a rabbitmq messaging queue and wait for messages to be dumped in a queue they are subscribed to. They will pick up these messages and do work accordingly. This might involve updating a db entry, scheduling a build or even spinning up a VM . You will want at least two workers running at any given time for HA. More for performance and/or capacity. Just some examples of these worker services are.

Queue Workers:

  - nova-conductor
  - nova-scheduler
  - nova-compute
  - nova-cert
  - heat-engine
  - cinder-scheduler
  - cinder-backup
  - cinder-volume (may need further research: https://blueprints.launchpad.net/cinder/+spec/cinder-volume-active-active-support) (https://review.openstack.org/#/c/327283/) The information on this one is sparse.

[Frank Kloeker (f-kloeker)]

created https://storyboard.openstack.org/#!/story/2005743