This project contains a set of filters for the Nova scheduler. These filters are intended to be used in conjunction with the default filters that ship with Nova.
The goal of those filters are to provide additional filtering capabilities that can be co-installed with any version of Nova until the features are upstreamed.
The failure domain filter is a filter that can be used to spread instances across different failure domains. A failure domain is a set of compute nodes that are expected to fail together. For example, compute nodes in the same rack or in the same power distribution unit (PDU) are expected to fail together.
Historically, some users have used the availability zone feature to spread instances across failure domains. However, the availability zone feature is not designed for this purpose and it has some limitations. There is also a need to spread instances across failure domains within the same availability zone.
The failure domain is defined as a metadata on host aggregates. The key of
the metadata is failure_domain. The value of the metadata is a string that
identifies the failure domain. The failure domain filter will spread instances
across different failure domains. It is up to the operator to define the
failure domains.