Map PCI pools to RP UUIDs

Nova's PCI scheduling (and the PCI claim) works based on PCI device
pools where the similar available PCI devices are assigned. The PCI
devices are now represented in placement as RPs. And the allocation
candidates during scheduling and the allocation after scheduling
now contain PCI devices. This information needs to affect the PCI
scheduling and PCI claim. To be able to do that we need to map PCI
device pools to RPs. We achieve that here by first mapping
PciDevice objects to RPs during placement PCI inventory reporting.
Then mapping pools to RPs based on the PCI devices assigned to the
pools.

Also because now ResourceTracker._update_to_placement() call updates
the PCI device pools the sequence of events needed to changed in the
ResourceTracker to:
1) run _update_to_placement()
2) copy the pools to the CompouteNode object
3) save the compute to the DB
4) save the PCI tracker

blueprint: pci-device-tracking-in-placement
Change-Id: I9bb450ac235ab72ff0d8078635e7a11c04ff6c1e

2 files changed, 22 insertions, 5 deletions

diff --git a/nova/compute/pci_placement_translator.py b/nova/compute/pci_placement_translator.py
index 8aa7a1f4a9..3ee52e303c 100644
--- a/nova/compute/pci_placement_translator.py
+++ b/nova/compute/pci_placement_translator.py
@@ -261,6 +261,12 @@ class PciResourceProvider:
         )
         provider_tree.update_traits(self.name, self.traits)
 
+        # Here we are sure the RP exists in the provider_tree. So, we can
+        # record the RP UUID in each PciDevice this RP represents
+        rp_uuid = provider_tree.data(self.name).uuid
+        for dev in self.devs:
+            dev.extra_info['rp_uuid'] = rp_uuid
+
     def update_allocations(
         self,
         allocations: dict,
@@ -598,6 +604,11 @@ def update_provider_tree_for_pci(
 
     pv.update_provider_tree(provider_tree)
     old_alloc = copy.deepcopy(allocations)
+    # update_provider_tree correlated the PciDevice objects with RPs in
+    # placement and recorded the RP UUID in the PciDevice object. We need to
+    # trigger an update on the device pools in the tracker to get the device
+    # RP UUID mapped to the device pools
+    pci_tracker.stats.populate_pools_metadata_from_assigned_devices()
     updated = pv.update_allocations(allocations, provider_tree)
 
     if updated:
diff --git a/nova/compute/resource_tracker.py b/nova/compute/resource_tracker.py
index ffbc7ed03f..ab60c77d96 100644
--- a/nova/compute/resource_tracker.py
+++ b/nova/compute/resource_tracker.py
@@ -991,8 +991,6 @@ class ResourceTracker(object):
         # notified when instances are deleted, we need remove all usages
         # from deleted instances.
         self.pci_tracker.clean_usage(instances, migrations)
-        dev_pools_obj = self.pci_tracker.stats.to_device_pools_obj()
-        cn.pci_device_pools = dev_pools_obj
 
         self._report_final_resource_view(nodename)
 
@@ -1314,13 +1312,23 @@ class ResourceTracker(object):
 
     def _update(self, context, compute_node, startup=False):
         """Update partial stats locally and populate them to Scheduler."""
+
+        self._update_to_placement(context, compute_node, startup)
+
+        if self.pci_tracker:
+            # sync PCI device pool state stored in the compute node with
+            # the actual state from the PCI tracker as we commit changes in
+            # the DB and in the PCI tracker below
+            dev_pools_obj = self.pci_tracker.stats.to_device_pools_obj()
+            compute_node.pci_device_pools = dev_pools_obj
+
         # _resource_change will update self.old_resources if it detects changes
         # but we want to restore those if compute_node.save() fails.
         nodename = compute_node.hypervisor_hostname
         old_compute = self.old_resources[nodename]
         if self._resource_change(compute_node):
             # If the compute_node's resource changed, update to DB. Note that
-            # _update_to_placement below does not supersede the need to do this
+            # _update_to_placement above does not supersede the need to do this
             # because there are stats-related fields in the ComputeNode object
             # which could have changed and still need to be reported to the
             # scheduler filters/weighers (which could be out of tree as well).
@@ -1333,8 +1341,6 @@ class ResourceTracker(object):
                 with excutils.save_and_reraise_exception(logger=LOG):
                     self.old_resources[nodename] = old_compute
 
-        self._update_to_placement(context, compute_node, startup)
-
         if self.pci_tracker:
             self.pci_tracker.save(context)