multiple key:value pairs in instance_type_extra_specs does not work

[David Kang]

 OS: CentOS 6.4
 Installation: using RDO Packstack.

 When one key:value pair is described in instance_type_extra_specs, the scheduler works fine.
However, when multiple key:value pairs are described in instance_type_extra_specs in /etc/nova/nova.conf, only one of them are properly filtered, but the other keys are recognized by None by the scheduler filter.
So, it always fails to find proper nova-compute.
I could use the same instance_type_extra_specs with Grizzly,
but it fails with Havana.

 Here is the flags that I used:

/etc/nova/nova.conf
instance_type_extra_specs=cpu_arch:x86_64, gpu_arch:fermi, gpus:2, hypervisor_type:LXC

flavor:
 nova flavor-list --extra-specs
+----+-----------+-----------+------+-----------+------+-------+-------------+-----------+----------------------------------------------------------------------------------------------------------+
| ID | Name | Memory_MB | Disk | Ephemeral | Swap | VCPUs | RXTX_Factor | Is_Public | extra_specs |
+----+-----------+-----------+------+-----------+------+-------+-------------+-----------+----------------------------------------------------------------------------------------------------------+

| 10 | cg1.small | 2048 | 10 | 0 | | 2 | 1.0 | True | {u'gpus': u'= 1', u'hypervisor_type': u's== LXC', u'gpu_arch': u's== fermi', u'cpu_arch': u's== x86_64'} |

[David Kang (dkang)]

It looks like that any key in instance_type_extra_specs field is not recognized by the compute_capabilities_filter at all.

For testing, I've added the following code into libvirt.HostState.update_status() function:

        data['pci_passthrough_devices'] = \
            self.driver.get_pci_passthrough_devices()

+ data["test_key"] = 'test_value'

Scheduler cannot find the value of the key from the state of compute node.

 Should the new key:value pair be inserted somewhere else?

[Jay Pipes (jaypipes)]

This is the commit that broke this, unless I am mistaken:

https://github.com/openstack/nova/commit/4b4f0d61836e527340511f6198c2f2c0bdf70888

[Jay Pipes (jaypipes)]

The reason I think the above commit (at least) is the issue is that after that commit, the HostManager.service_states dict existed but no longer was used by the scheduler, and therefore no capabilities were extended on the HostState objects passed to the filter's host_passes() method.

[Guangya Liu (Jay Lau) (jay-lau-513)]

Can you please append more for your case? Which filter are you using?

Also curious do we support adding instance_type_extra_specs to nova.conf? Who will use it?

/etc/nova/nova.conf
instance_type_extra_specs=cpu_arch:x86_64, gpu_arch:fermi, gpus:2, hypervisor_type:LXC

[Guangya Liu (Jay Lau) (jay-lau-513)]

Just notice that seems this parameter was only for bare-metal driver: https://github.com/openstack/nova/blob/master/nova/virt/baremetal/driver.py#L57 and also the parameter has been renamed to flavor_extra_specs

@Jay Pipes, this commit https://github.com/openstack/nova/commit/4b4f0d61836e527340511f6198c2f2c0bdf70888 was merged in 2014-01-28 and this bug was reported in 2013-10-28, so I think this is not related to the patch, right? ;-)

@David, hope I can get more input here for your case. Thanks.

[David Kang (dkang)]

About the scheduler filter (sorry for my late relpy), I believe I've used ComputeCapabilitiesFilter.

About the usage:

Actually we've used the flag to specify heterogeneous architecture support. In addition to x86 architecture, we wanted to support other architecture using the instance_type_extra_specs flag.
And we put dynamically changing values in the flag.
For example,

instance_type_extra_specs=cpu_arch:x86_64, gpu_arch:fermi, gpus:2, hypervisor_type:LXC

shows that the nova-compute node has two GPUs of NVdia fermi and runs LXC hypervisor.
The 'gpus' key is dynamically changing at run-time as gpus are allocated and deallocated.
Most static keys can be handled by host-aggregate method.
But keys with dynamic values cannot be handled by host-aggregate method.

Even before the merge of "https://github.com/openstack/nova/commit/4b4f0d61836e527340511f6198c2f2c0bdf70888 merged in 2014-01-28", the dynamic update mechanism was broken.

[Mikyung Kang (mkkang)]

So, ISI tried to fix this bug for Havana and Icehouse releases as follows, reviewers decided to deprecate this capability:

https://review.openstack.org/#/c/62088/

[Justin Shepherd (jshepher)]

This capability was marked as deprecated in https://review.openstack.org/#/c/62088/