Resource tracker double counts vcpus_used

The hypervisor is returning its own view of the number of vcpus used. This is copied to the compute_node record and is not reset before adding the resource tracker view. Code to reset the memory and disk is included in _update_usage_from_instances() as below. The extensible resource plugins are called here as well. Now the vcpu resource has gone and the accounting is directly in the compute_node object the vcpus_used should be reset to 0 here.

    def _update_usage_from_instances(self, context, instances):
        """Calculate resource usage based on instance utilization. This is
        different than the hypervisor's view as it will account for all
        instances assigned to the local compute host, even if they are not
        currently powered on.
        """
        self.tracked_instances.clear()

        # set some initial values, reserve room for host/hypervisor:
        self.compute_node.local_gb_used = CONF.reserved_host_disk_mb / 1024
        self.compute_node.memory_mb_used = CONF.reserved_host_memory_mb
        self.compute_node.free_ram_mb = (self.compute_node.memory_mb -
                                         self.compute_node.memory_mb_used)
        self.compute_node.free_disk_gb = (self.compute_node.local_gb -
                                          self.compute_node.local_gb_used)
        self.compute_node.current_workload = 0
        self.compute_node.running_vms = 0

        # Reset values for extended resources
        self.ext_resources_handler.reset_resources(self.compute_node,
                                                   self.driver)

Fix proposed to branch: master
Review: https://review.openstack.org/248742

Reviewed: https://review.openstack.org/248742
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=69eca4c60d762b673fafb13dd12f285d7f1f1239
Submitter: Jenkins
Branch: master

commit 69eca4c60d762b673fafb13dd12f285d7f1f1239
Author: Paul Murray <email address hidden>
Date: Mon Nov 23 14:50:14 2015 +0000

    Fix resource tracker VCPU counting

    The resource track is double counting the number of
    vcpus used because it counts the virt driver's view
    and the resource tracker's view. It should reset
    the count to 0 before counting the resourcet racker's
    view. this is done for other resources (memory, disk)
    in _update_usage_from_instances().

    This was not caught in unit tests because they always
    provide a virt driver view with no resource used for
    tests that assume instances are already running.

    This patch adds the reset to 0 and modifies the
    tests to provide correct counts of used resources
    from the virt driver.

    Change-Id: I21acfd7d5b5ed442cd1e9049c986a06bd2b59b81
    Closes-Bug: #1519001

This issue was fixed in the openstack/nova 13.0.0.0b1 development milestone.