CLOUDSTACK-10188 - Resource Accounting for primary storage is Broken when Domains are in use (#2362)

During storage expunge domain resource statistics for primary storage space resource counter is not updated for domain. This leads to the situation when domain resource statistics for primary storage is overfilled (statistics only increase but not decrease).

Global scheduled task resourcecount.check.interval > 0 provides a workaround but not fixes the problem truly because when accounts inside domains use primary_storage allocation/deallocation intensively it leads to service block of operation.

NB: Unable to implement marvin tests because it (marvin) places in database weird primary storage volume size of 100 when creating VM from template. It might be a sign of opening a new issue for that bug.

Author: bwsw

api/src/com/cloud/user/ResourceLimitService.java

@@ -28,7 +28,7 @@
 public interface ResourceLimitService {
 
     static final ConfigKey<Long> ResourceCountCheckInterval = new ConfigKey<Long>("Advanced", Long.class, "resourcecount.check.interval", "300",
-            "Time (in seconds) to wait before retrying resource count check task. Default is 300, Setting this to 0 will not run the task", false);
+            "Time (in seconds) to wait before running resource recalculation and fixing task. Default is 300 seconds, Setting this to 0 disables execution of the task", false);
 
     /**
      * Updates an existing resource limit with the specified details. If a limit doesn't exist, will create one.

engine/orchestration/src/org/apache/cloudstack/engine/orchestration/VolumeOrchestrator.java

@@ -1576,8 +1576,7 @@ public void destroyVolume(Volume volume) {
             UsageEventUtils.publishUsageEvent(EventTypes.EVENT_VOLUME_DELETE, volume.getAccountId(), volume.getDataCenterId(), volume.getId(), volume.getName(),
                     Volume.class.getName(), volume.getUuid(), volume.isDisplayVolume());
             _resourceLimitMgr.decrementResourceCount(volume.getAccountId(), ResourceType.volume, volume.isDisplay());
-            //FIXME - why recalculate and not decrement
-            _resourceLimitMgr.recalculateResourceCount(volume.getAccountId(), volume.getDomainId(), ResourceType.primary_storage.getOrdinal());
+            _resourceLimitMgr.decrementResourceCount(volume.getAccountId(), ResourceType.primary_storage, volume.isDisplay(), new Long(volume.getSize()));
         } catch (Exception e) {
             s_logger.debug("Failed to destroy volume" + volume.getId(), e);
             throw new CloudRuntimeException("Failed to destroy volume" + volume.getId(), e);