Hi, Alonso:

 

I'm not sure that the configuration had been changed. You can check the nova compute's log (you need enable debug mode), search 'cpu_models', you'll see a record similar as below(the record should printed when service start):

 

" 2019-11-08 18:21:58.966 6 DEBUG oslo_service.service [req-16a54cb4-1d45-4119-9fd4-6d84f471ebb0 - - - - -] libvirt.cpu_models = [SandyBridge,IvyBridge,Haswell,Broadwell] log_opt_values /var/lib/kolla/venv/lib/python2.7/site-packages/oslo_config/cfg.py:2585"

 

Please check the list, if it’s same as you set, then the configuration changed success.

 

Best Regards

 

On 2019/11/7, 4:19 AM, "Alonso, Juan Carlos" <starlingx-discuss-bounces+wang.ya=99cloud.net@lists.starlingx.io on behalf of juan.carlos.alonso@intel.com> wrote:

 

    Hello Ya,

   

    I have an standard configuration local storage (2+2) baremetal environment, I can access to nova-compute service for both computes but I did not found a list of 'cpu_models', just found 'cpu_mode' and it is not set as 'custom' as documentation says:

    [libvirt]

    cpu_mode = host-model

   

    So, I executed:

    $ system helm-override-update --set livbirt.cpu_mode=custom stx-openstack nova openstack

    $ system helm-override-update --set cpu_models=SandyBridge,IvyBridge,Haswell,Broadwell stx-openstack nova openstack

   

    I check again /etc/nova/nova.conf inside pod, and still saw just 'cpu_mode = host-model'

    Is this configuration be modified? If yes, how can I do it?

   

    Also I could create flavors and instances with below configurations:

    trait:HW_CPU_X86_AVX='required'

    trait:HW_CPU_X86_AVX2='required'

    trait:HW_CPU_X86_AVX512F='required'

   

    One instance for each configuration and one instance with those three configuration in a same flavor.

    But, how can I check the CPU model in the instance?

   

    In the test cases I mention at the beginning of this threat it says that the instances with flavor property trait:HW_CPU_X86_AVX512F='required' should get ERROR, but it could be launched as ACTIVE.

   

    Regards.

    Juan Carlos Alonso

   

    -----Original Message-----

    From: wang.ya <wang.ya@99cloud.net>

    Sent: Monday, November 4, 2019 9:01 PM

    To: Hu, Yong <yong.hu@intel.com>; Jones, Bruce E <bruce.e.jones@intel.com>; Alonso, Juan Carlos <juan.carlos.alonso@intel.com>; Troyer, Dean <dean.troyer@intel.com>

    Cc: Cabrales, Ada <ada.cabrales@intel.com>

    Subject: Re: Patch Elimination Train version

   

    Hi, yong:

   

    *vCPU Models selection* feature has been completed in train release.

   

    The feature's workflow is as follow:

   

    1. Configure the *cpu_models* option under [libvirt] module of nova-compute service.

         The *cpu_models* option is an ordered list of CPU models. It's sorted by CPU flags, from less to more. For example:

               kvm64, IvyBridge, Haswell

         In above example, kvm64 only has base flags, IvyBridge has flag *avx*, Haswell have both *avx* and *avx2*.

         Please note that the configured CPU models must compatible with compute node's host CPU.

   

    2. Create/Set flavor with required traits. The nova compute will select the CPU model from [libvirt]/cpu_models which first meet the flavor's requested flags. For example:

               openstack flavor set 1 --property trait:HW_CPU_X86_AVX=required

         In above example, we mark the flavor request flag *avx*, if the CPU of compute node has flag *avx*, then the instance can be scheduled success and created with CPU model IvyBridge, otherwise the API call will get an error: "No valid host".

    

    Spec: http://specs.openstack.org/openstack/nova-specs/specs/train/approved/cpu-model-selection.html

    Code: https://review.opendev.org/#/q/status:merged+project:openstack/nova+branch:master+topic:bp/cpu-model-selection

   

    Hope the above info can help you. Also, I'll online in today's meeting :)

   

    Best Regards

   

    

    On 2019/11/5, 9:13 AM, "Yong Hu" <yong.hu@intel.com> wrote:

   

        + Wang Ya.

       

        Hi Ya,

        I remember your team has been upstreaming the related patches.

        Could you share some info with us here or in the meeting tonight?

       

        regards,

        Yong

       

        On 2019/11/5 6:37 AM, Jones, Bruce E wrote:

        > + Dean and Yong (the new PL for the OpenStack team)

        >

        > AFAIK the CPU model changes that went into Train are internal only and

        > don’t impact OpenStack operation.

        >

        > Suggest you bring all of this to the distro.openstack meeting tomorrow

        > where it can be discussed with the right people.

        >

        >       brucej

        >

        > *From:* Alonso, Juan Carlos

        > *Sent:* Monday, November 4, 2019 2:30 PM

        > *To:* Jones, Bruce E <bruce.e.jones@intel.com>

        > *Cc:* Cabrales, Ada <ada.cabrales@intel.com>

        > *Subject:* Patch Elimination Train version

        >

        > Hi Bruce,

        >

        > In last test plan for Patch Elimination Stein version, there was some

        > *vCPU Model (Nova)* feature test cases that was marked as Unlikely status:

        >

        > 1. Create a flavor without any extra specs. Launch instance using this

        > flavor. Check instance cpu_model is 'kvm64'

        >

        > 2. Create a flavor with extra specs: 'trait:HW_CPU_X86_AVX2=required'.

        > Launch instance using this flavor. Check instance is Haswell and is in

        > compute node 2.

        >

        > 3. Create a flavor with extra specs:

        > 'trait:HW_CPU_X86_AVX512F=required'. Launch instance using this flavor.

        > Check API call get an error because no compute node has property 'avx512F'.

        >

        > 4. Create a flavor with multiple extra specs: 'trait:xxx'. Launch

        > instance using this flavor.

        >

        > 5. Create a flavor and ignoring cpu_model. Launch instance using this

        > flavor.

        >

        > For Train version, is such *vCPU Model (Nova)* feature still out?

        >

        > Also, there was some Live Migration Bugs table related to Neutron:

        >

        > Enable delete bound trunk for linux bridge agent

        >

        > 0040-rpc-removing-timeout-backoff-multiplier.patch (2e4ca00)

        >

        > 0062-dvr-force-admin-state-update-before-distri.patch (edf9731)

        >

        > 0065-dvr-do-not-create-agent-gateway-ports-unle.patch (2857911)

        >

        > 0077-api-reject-routes-with-invalid-network-val.patch (6ed8fe6)

        >

        > 0014-add-support-for-querying-quotas-with-usage.patch (71c07d7)

        >

        > Prevent DHCP from processing stale RPC messages

        >

        > 0039-dhcp-handle-concurrent-port-creation-error.patch (a38f89)

        >

        > 0047-wsgi-prevent-accepting-socket-without-a-gr.patch (8e72491)

        >

        > I need to know if such features/tests are going to be testes or not, can

        > you help me to verify?

        >

        > Test Plan for Stein:

        > https://docs.google.com/spreadsheets/d/1jsTyzy1HT7VR4EUsbBRaTYvs0BlXeGWkhLD0pJmFChc/edit#gid=1511025137

        >

        > Test Plan for Train:

        > https://docs.google.com/spreadsheets/d/1n9P7as4iqO55DytIRuIGdyc7x0n2WM_q-cMfHcItJlY/edit?pli=1#gid=184151983

        >

        > Regards.

        >

        > Juan Carlos Alonso

        >

        

        

    

    

    _______________________________________________

    Starlingx-discuss mailing list

    Starlingx-discuss@lists.starlingx.io

    http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss