Wrong hypervisor statistics reported

[Liyingjun]

The hypervisor statistics reported through nova hypervisor-stats is wrong if there are other nova services running on the compute node, output below:

$ nova hypervisor-stats
+----------------------+-------+
| Property | Value |
+----------------------+-------+
| count | 5 |
| current_workload | 0 |
| disk_available_least | 395 |
| free_disk_gb | 455 |
| free_ram_mb | 34160 |
| local_gb | 455 |
| local_gb_used | 0 |
| memory_mb | 39920 |
| memory_mb_used | 5760 |
| running_vms | 15 |
| vcpus | 20 |
| vcpus_used | 15 |
+----------------------+-------+

But i only have 3 instances launched, and 1 hypervisor with 4 vcpus... seems like the result is multiplied by 5.

$ nova hypervisor-list
+----+---------------------+
| ID | Hypervisor hostname |
+----+---------------------+
| 1 | holly |
+----+---------------------+

$ nova hypervisor-show 1
+---------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Property | Value |
+---------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| cpu_info_arch | x86_64 |
| cpu_info_features | ["pge", "clflush", "sep", "syscall", "vme", "dtes64", "tsc", "xsave", "vmx", "xtpr", "cmov", "ssse3", "est", "pat", "monitor", "lm", "msr", "nx", "fxsr", "tm", "sse4.1", "pae", "acpi", "de", "mmx", "osxsave", "cx8", "mce", "mtrr", "ht", "pse", "lahf_lm", "pdcm", "mca", "apic", "sse", "ds", "pni", "tm2", "sse2", "ss", "pbe", "fpu", "cx16", "pse36", "ds_cpl"] |
| cpu_info_model | Penryn |
| cpu_info_topology_cores | 4 |
| cpu_info_topology_sockets | 1 |
| cpu_info_topology_threads | 1 |
| cpu_info_vendor | Intel |
| current_workload | 0 |
| disk_available_least | 79 |
| free_disk_gb | 91 |
| free_ram_mb | 6832 |
| host_ip | 192.168.70.11 |
| hypervisor_hostname | holly |
| hypervisor_type | QEMU |
| hypervisor_version | 2000000 |
| id | 1 |
| local_gb | 91 |
| local_gb_used | 0 |
| memory_mb | 7984 |
| memory_mb_used | 1152 |
| running_vms | 3 |
| service_disabled_reason | - |
| service_host | holly |
| service_id | 5 |
| state | up |
| status | enabled |
| vcpus | 4 |
| vcpus_used | 3 |
+---------------------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/152372

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/152372
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=968275f1efe27821fab0b949353ddd7e6fcec583
Submitter: Jenkins
Branch: master

commit 968275f1efe27821fab0b949353ddd7e6fcec583
Author: liyingjun <email address hidden>
Date: Tue Feb 3 10:11:29 2015 +0800

    Fixes logic in compute_node_statistics

    The filter is wrong in compute_node_statistics, currently, if there
    are other nova services in the compute node, the statistics will be
    caculated multiplied by services count, because it didn't filter
    out the nova-compute services. Fix the bug by adding a filter
    'models.Service.binary == "nova-compute"' in query.

    Change-Id: Icc16e92e61594e8700608006f56c45d424d1edf9
    Closes-bug: #1417340