Nejc Saje reported on the openstack-dev list that our hash ring was not
stable as hosts were added or removed. This is a significant issue since
the hash ring is meant to minimize the impact of adding or removing
conductors to only approximately the change in capacity.
IMPORTANT: When this patch is added to a running Ironic environment it
will cause ring calculations to change. This will cause most existing
node -> conductor mappings to be invalid, which will cause existing
deployed nodes to be unable to reboot (if using a netbooting driver like
pxe_ipmi). The cluster will self-correct over time as the periodic task
_sync_hash_ring() rebuilds PXE environments.
NOTE: The meaning of the hash_partition_exponent configuration setting
had to be changed - if we define the total number of partitions of the
ring via it then adding a host simultaneously removes some partitions of
each existing host as well as adding new partitions from the new host -
this results in more perturbation. Instead we use it to define the
number of partitions per host. Existing deployments will continue to
work, but may be able to reduce some CPU overhead by shrinking their
setting.
Change-Id: Iad3c9f6b460e9162ca91d5719e42f7950f2f18df
Closes-Bug: #1365334
DocImpact: See above.
UpgradeImpact: See above.
b272e61218