Fix XenAPI performance issue

This patch implements 'list_instance_uuids' in the xenapi virt driver so
that compute manager's '_get_instances_on_driver' can operate more
efficiently.

Fixes bug 1097980

The cleanup_running_deleted_instances periodic task uses the above call
while context has been modified to be read_deleted='yes'.  Without
list_instance_uuids being implemented in xenapi, there's a fallback to
querying all instances on the host.  Because of read_deleted='yes', this
queries all instances that have ever lived on the host.  In a very busy
environment where instances are repeatedly built and destroyed, one can
end up with thousands of deleted instances.  Now that we are storing
instance_type data in system_metadata and system_metadata is joined with
every instance_get, this results in 10x the number of rows being
returned with sqlalchemy... the fallback doesn't perform well enough.

Change-Id: I4bbfd69c9769807cec813af757665f03d9643460

0 files changed, 0 insertions, 0 deletions