Large scale operators tend to have needs that involve creating well defined and delinated resources. In some cases, these systems may reside close by or in far away locations. Reasoning may be simple or complex, and yet is only known to the deployer and operator of the infrastructure.
A common case is the need for delineated high availability domains where it would be much more efficient to manage a datacenter in Antarctica with a conductor in Antarctica, as opposed to a conductor in New York City.
How it works¶
Starting in ironic 11.1, each node has a
conductor_group field which
influences how the ironic conductor calculates (and thus allocates)
baremetal nodes under ironic’s management. This calculation is performed
independently by each operating conductor and as such if a conductor has
[conductor]conductor_group configuration option defined in its
ironic.conf configuration file, the conductor will then be limited to
only managing nodes with a matching
Any conductor without a
[conductor]conductor_group setting will
only manage baremetal nodes without a
conductor_group value set upon
node creation. If no such conductor is present when conductor groups are
configured, node creation will fail unless a
specified upon node creation.
Nodes without a
conductor_group setting can only be managed when a
conductor exists that does not have a
defined. If all conductors have been migrated to use a conductor group,
such nodes are effectively “orphaned”.
How to use¶
A conductor group value may be any case insensitive string up to 255
characters long which matches the
[conductor]conductor_groupoption in ironic.conf on one or more, but not all conductors:
[conductor] conductor_group = OperatorDefinedString
Restart the ironic-conductor service.
Set the conductor group on one or more nodes:
baremetal node set \ --conductor-group "OperatorDefinedString" <uuid>
As desired and as needed, remaining conductors can be updated with the first two steps. Please be mindful of the constraints covered earlier in the document related to ability to manage nodes.