An overview of the Freezer-DR (Disaster Recovery) project

Freezer-DR is a project that provides compute node high availability for OpenStack. The Freezer-DR comprises 4 main boxes:

  • Monitors: collecting data from compute nodes
  • Fencers: do fencing operations
  • Evacuators: to evacuate VMs
  • Notifiers: notification

Freezer-DR is designed using an extensible architecture which allows us to plug any software solution/library into the above boxes by implementing the drivers. For example, I can use Monasca as the monitor, IPMI or libvirt as fencers, open stack nova client as evacuator, python SMTP lib as the notifier, as long as I have the drivers for those software or libraries. Right now, there is only a limited number of supported drivers for monitors, fencers, evacuators, and notifiers but you can always easily write your own drivers to use the desired libraries or software.

Below figure describes the workflow of Freezer-DR:

(1) Freezer-DR monitors all the compute nodes
(2) If it found any fail node, Freezer-DR will fence that node
(3) It will then try to evacuate all the VMs of the failed node to another node.
(4) Freezer-DR notifies the user.

Github repository: