<div dir="ltr"><div class="gmail_default" style="font-family:tahoma,sans-serif">I made some changes and now I have a implementation that is compatible with Vishy's implementation, just had to add one change to Nova. </div>

<div class="gmail_default" style="font-family:tahoma,sans-serif">Please give your feedbacks. As for id's or names, im storing names in keystone and sending ids to nova.</div><div class="gmail_default" style="font-family:tahoma,sans-serif">

<br></div><div class="gmail_default"><font face="tahoma, sans-serif">link to github: <a href="https://github.com/tellesnobrega/keystone_hierarchical_projects">https://github.com/tellesnobrega/keystone_hierarchical_projects</a></font><br>

</div><div class="gmail_default"><font face="tahoma, sans-serif"><br></font></div><div class="gmail_default"><font face="tahoma, sans-serif">Thanks</font></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">

On Thu, Feb 20, 2014 at 7:18 PM, Vishvananda Ishaya <span dir="ltr"><<a href="mailto:vishvananda@gmail.com" target="_blank">vishvananda@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">

<div style="word-wrap:break-word"><br><div><div>On Feb 19, 2014, at 5:58 PM, Adam Young <<a href="mailto:ayoung@redhat.com" target="_blank">ayoung@redhat.com</a>> wrote:</div><br><blockquote type="cite">
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    <div>On 02/18/2014 02:28 PM, Vishvananda
      Ishaya wrote:<br>
    </div>
    <blockquote type="cite">
      
      <br>
      <div>
        <div>On Feb 18, 2014, at 11:04 AM, Adam Young <<a href="mailto:ayoung@redhat.com" target="_blank">ayoung@redhat.com</a>>
          wrote:</div>
        <br>
        <blockquote type="cite">
          
          <div bgcolor="#FFFFFF" text="#000000">
            <div>On 02/18/2014 12:53 PM, Telles
              Nobrega wrote:<br>
            </div>
            <blockquote type="cite">
              <div dir="ltr">
                <div class="gmail_default" style="font-family:tahoma,sans-serif">Hello everyone,</div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"> Me and Raildo
                  were responsible to implement Hierarchical Projects in
                  Keystone.</div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"> Here is our
                  first prototype: <a href="https://github.com/tellesnobrega/keystone_hierarchical_projects" target="_blank">https://github.com/tellesnobrega/keystone_hierarchical_projects</a></div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"> <br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif">We want to have
                  it tested with Vishy's implementation this week.<br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif">Here is a  guide
                  on how to test the implementation:<br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"><br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif"> 1. Start a
                  devstack using the keystone code;</div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif">2. Create a new
                  project using the following body:</div>
                <div class="gmail_default"><span style="font-family:tahoma,sans-serif">    </span><font face="tahoma, sans-serif">{</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">    "project": {</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">        "description": "test_project",</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">        "domain_id": "default",</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">        "parent_project_id":
                    "$parent_project_id",</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">        "enabled": true,</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">        "name": "test_project"</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">    }</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">}</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif"><br>
                  </font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">3. Give an user a role in the project;</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">4. Get a token for "test_project" and
                    check that the hierarchy is there like the
                    following:</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">    </font><span>{</span></div>
                <pre style="margin-top:0px;margin-bottom:0px;padding:5px 0px;font-family:'Bitstream Vera Sans Mono',monospace;font-size:13px">    "token": {
        "methods": [
            "password"
        ],
        "roles": [
            {
                "id": "c60f0d7461354749ae8ac8bace3e35c5",
                "name": "admin"
            }
        ],
        "expires_at": "2014-02-18T15:52:03.499433Z",
        "project": {
            "hierarchical_ids": "<div class="gmail_default" style="font-family:tahoma,sans-serif;display:inline">openstack.</div>8a4ebcf44ebc47e0b98d3d5780c1f71a.de2a7135b01344cd82a02117c005ce47",</pre>


              </div>
            </blockquote>
            <br>
            These should be names, not Ids.  There is going to be a need
            to move projecst around inside the hierarchy, and the ID
            stays the same.  Lets get this right up front.<br>
          </div>
        </blockquote>
        <div><br>
        </div>
        Can you give more detail here? I can see arguments for both ways
        of doing this but continuing to use ids for ownership is an
        easier choice. Here is my thinking:</div>
      <div><br>
      </div>
      <div>1. all of the projects use ids for ownership currently so it
        is a smaller change</div>
    </blockquote>
    That does not change.  It is the hierarchy that is labeled by name.<br></div></blockquote><div><br></div>The issue is that we are storing the hierarchy of ownership in nova. We can either store the hierarchy by id or by name. Note that we are not adding a new field for this hierarchy but using the existing ownership field (which is called project_id in nova). My point is that if we use ids, then this field would be backwards compatible. If we decide to use name instead (which has some advantages for display purposes), then we would need some kind of db sync migration which modifies all of the fields from id -> name.<br>

<blockquote type="cite"><div bgcolor="#FFFFFF" text="#000000">
    <br>
    <blockquote type="cite">
      <div>2. renaming a project in keystone would not invalidate the
        ownership hierarchy (Note that moving a project around would
        invalidate the hierarchy in both cases)</div>
      <div><br>
      </div>
    </blockquote>
    Renaming would not change anything. <br>
    <br>
    I would say the rule should be this:  Ids are basically uuids, and
    are immutable.  Names a mutable.  Each project has a parent Id.  A
    project can either be referenced directly by ID, oir hierarchically
    by name.  In addition, you can navigate to a project by traversing
    the set of ids, but you need to know where you are going.  THus the
    array <br>
    <br>
    ['abcd1234',fedd3213','3e3e3e3e'] would be a way to find a project,
    but the project ID for the lead node would still be just '3e3e3e3e’.<br></div></blockquote><div><br></div>As I mention above, all of this makes sense inside of keystone, but doesn’t address the problem of how we are storing the hierarchy on the nova side. The owner field in nova can be:</div>

<div><br></div><div>1) abcd1234.fedd3213.3e3e3e3e</div><div><br></div><div>or it can be:</div><div><br></div><div>2) orga.proja.suba</div><div><br></div><div>To explicitly state the tradeoffs</div><div><br></div><div>1 is backwards compatible +</div>

<div>1 doesn’t need to be updated if a project is renamed +</div><div>1 is not user friendly (need to map ids to names to display to the user) -</div><div><br></div><div>both need to be updated if a project is moved in the hierarchy</div>

<div><br></div><div>Vish</div><div><br><blockquote type="cite"><div bgcolor="#FFFFFF" text="#000000">
    <br>
    <br>
    <blockquote type="cite">
      <div>OTOTH the advantage of names is it makes displaying the
        ownership much easier on the service side.</div>
      <div><br>
      </div>
      <div>Vish</div>
      <div><br>
      </div>
      <div>
        <blockquote type="cite">
          <div bgcolor="#FFFFFF" text="#000000"> <br>
            <blockquote type="cite">
              <div dir="ltr">
                <pre style="margin-top:0px;margin-bottom:0px;padding:5px 0px;font-family:'Bitstream Vera Sans Mono',monospace;font-size:13px">            "hierarchy": "test1",
            "domain": {
                "id": "default",
                "name": "Default"
            },
            "id": "de2a7135b01344cd82a02117c005ce47",
            "name": "test1"
        },
        "extras": {},
        "user": {
            "domain": {
                "id": "default",
                "name": "Default"
            },
            "id": "895864161f1e4beaae42d9392ec105c8",
            "name": "admin"
        },
        "issued_at": "2014-02-18T14:52:03.499478Z"
    }
}</pre>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif"><br>
                  </font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif">Openstack is the root project of the
                    tree, it can be seen also when getting a token for
                    the admin project or other default project in
                    Devstack.</font></div>
                <div class="gmail_default"><font face="tahoma,
                    sans-serif"><br>
                  </font></div>
                <div class="gmail_default"><span style="font-family:tahoma,sans-serif">Hope to hear
                    your feedbacks soon.</span><font face="tahoma,
                    sans-serif"><br>
                  </font></div>
                <div class="gmail_default"><br>
                </div>
                <div class="gmail_default" style="font-family:tahoma,sans-serif">Thanks</div>
              </div>
              <div class="gmail_extra"><br>
                <br>
                <div class="gmail_quote">On Mon, Feb 17, 2014 at 6:09
                  AM, Vinod Kumar Boppanna <span dir="ltr"><<a href="mailto:vinod.kumar.boppanna@cern.ch" target="_blank">vinod.kumar.boppanna@cern.ch</a>></span>
                  wrote:<br>
                  <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Dear
                    Vish,<br>
                    <br>
                    I will change the concept of parsing roles upto leaf
                    node to parsing the roles to the top upto the level
                    1. But i have small doubt and i want to confirm with
                    you before doing this change.<br>
                    <br>
                    If there are lets say 10 levels in the hierarchy and
                    the user is getting authenticated at level 9. Should
                    i check the roles starting from level 9 upto level
                    1. Ofcourse, the difference here is (compared to
                    what i put in the wiki page) that, only roles at
                    each level (if different) needs to be added to scope
                    and no need of adding the project name and role
                    individually. Is this ok, considering the fact that
                    the more deeper in the hierarchy the user is getting
                    authenticated, the more time needed to parse upto
                    the level 1.<br>
                    <br>
                    I will wait for your response and then modify the
                    POC accordingly.<br>
                    <div><br>
                      Thanks & Regards,<br>
                      Vinod Kumar Boppanna<br>
                    </div>
                    ________________________________________<br>
                    From: <a href="mailto:openstack-dev-request@lists.openstack.org" target="_blank">openstack-dev-request@lists.openstack.org</a>
                    [<a href="mailto:openstack-dev-request@lists.openstack.org" target="_blank">openstack-dev-request@lists.openstack.org</a>]<br>
                    Sent: 16 February 2014 22:21<br>
                    To: <a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a><br>
                    Subject: OpenStack-dev Digest, Vol 22, Issue 45<br>
                    <div><br>
                      Send OpenStack-dev mailing list submissions to<br>
                              <a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a><br>
                      <br>
                      To subscribe or unsubscribe via the World Wide
                      Web, visit<br>
                              <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      or, via email, send a message with subject or body
                      'help' to<br>
                              <a href="mailto:openstack-dev-request@lists.openstack.org" target="_blank">openstack-dev-request@lists.openstack.org</a><br>
                      <br>
                      You can reach the person managing the list at<br>
                              <a href="mailto:openstack-dev-owner@lists.openstack.org" target="_blank">openstack-dev-owner@lists.openstack.org</a><br>
                      <br>
                      When replying, please edit your Subject line so it
                      is more specific<br>
                      than "Re: Contents of OpenStack-dev digest..."<br>
                      <br>
                      <br>
                      Today's Topics:<br>
                      <br>
                    </div>
                       1. Re: [Nova][VMWare] VMwareVCDriver related to
                    resize/cold<br>
                          migration (Gary Kotton)<br>
                       2. [Neutron]Do you think tanent_id should be
                    verified (Dong Liu)<br>
                       3. Re: [Nova][VMWare] VMwareVCDriver related to
                    resize/cold<br>
                          migration (Jay Lau)<br>
                       4. [neutron][policy] Using network services with
                    network<br>
                          policies (Mohammad Banikazemi)<br>
                       5. Re: [Nova][VMWare] VMwareVCDriver related to
                    resize/cold<br>
                          migration (Jay Lau)<br>
                       6. Re: [keystone] role of Domain in VPC
                    definition (Harshad Nakil)<br>
                       7. Re: VPC Proposal (Harshad Nakil)<br>
                       8. Re: VPC Proposal (Allamaraju, Subbu)<br>
                       9. Re: VPC Proposal (Harshad Nakil)<br>
                      10. Re: VPC Proposal (Martin, JC)<br>
                      11. Re: [keystone] role of Domain in VPC
                    definition<br>
                          (Allamaraju, Subbu)<br>
                      12. Re: [keystone] role of Domain in VPC
                    definition (Harshad Nakil)<br>
                      13. Re: [keystone] role of Domain in VPC
                    definition<br>
                          (Allamaraju, Subbu)<br>
                      14. Re: [OpenStack-Infra] [TripleO] promoting
                    devtest_seed and<br>
                          devtest_undercloud to voting, + experimental
                    queue for<br>
                          nova/neutron etc. (Robert Collins)<br>
                      15. Re: [OpenStack-Infra] [TripleO] promoting
                    devtest_seed and<br>
                          devtest_undercloud to voting, + experimental
                    queue for<br>
                          nova/neutron etc. (Robert Collins)<br>
                      16. Re: [keystone] role of Domain in VPC
                    definition (Ravi Chunduru)<br>
                      17. Re: VPC Proposal (Ravi Chunduru)<br>
                      18. Re: OpenStack-dev Digest, Vol 22, Issue 39
                    (Vishvananda Ishaya)<br>
                      19. Re: heat run_tests.sh fails with one huge line
                       of      output<br>
                          (Mike Spreitzer)<br>
                    <br>
                    <br>
----------------------------------------------------------------------<br>
                    <br>
                    Message: 1<br>
                    Date: Sun, 16 Feb 2014 05:40:05 -0800<br>
                    From: Gary Kotton <<a href="mailto:gkotton@vmware.com" target="_blank">gkotton@vmware.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to<br>
                            resize/cold migration<br>
                    Message-ID: <<a href="mailto:CF268BE4.465C7%25gkotton@vmware.com" target="_blank">CF268BE4.465C7%gkotton@vmware.com</a>><br>
                    <div>Content-Type: text/plain;
                      charset="us-ascii"<br>
                      <br>
                      Hi,<br>
                    </div>
                    There are two issues here.<br>
                    The first is a bug fix that is in review:<br>
                    - <a href="https://review.openstack.org/#/c/69209/" target="_blank">https://review.openstack.org/#/c/69209/</a>
                    (this is where they have the same configuration)<br>
                    The second is WIP:<br>
                    - <a href="https://review.openstack.org/#/c/69262/" target="_blank">https://review.openstack.org/#/c/69262/</a>
                    (we need to restore)<br>
                    Thanks<br>
                    Gary<br>
                    <br>
                    From: Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a><mailto:<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>>><br>


                    Reply-To: "OpenStack Development Mailing List (not
                    for usage questions)" <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a><mailto:<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>>><br>


                    Date: Sunday, February 16, 2014 6:39 AM<br>
                    To: OpenStack Development Mailing List <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a><mailto:<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>>><br>


                    Subject: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to resize/cold migration<br>
                    <br>
                    Hey,<br>
                    <br>
                    I have one question related with OpenStack
                    vmwareapi.VMwareVCDriver resize/cold migration.<br>
                    <br>
                    The following is my configuration:<br>
                    <br>
                     DC<br>
                        |<br>
                        |----Cluster1<br>
                        |          |<br>
                        |          |----9.111.249.56<br>
                        |<br>
                        |----Cluster2<br>
                                   |<br>
                                   |----9.111.249.49<br>
                    <br>
                    Scenario 1:<br>
                    I started two nova computes manage the two clusters:<br>
                    1) nova-compute1.conf<br>
                    cluster_name=Cluster1<br>
                    <br>
                    2) nova-compute2.conf<br>
                    cluster_name=Cluster2<br>
                    <br>
                    3) Start up two nova computes on host1 and host2
                    separately<br>
                    4) Create one VM instance and the VM instance was
                    booted on Cluster2 node  9.111.249.49<br>
                    | OS-EXT-SRV-ATTR:host                 | host2 |<br>
                    | OS-EXT-SRV-ATTR:hypervisor_hostname  |
                    domain-c16(Cluster2)                                
                        |<br>
                    5) Cold migrate the VM instance<br>
                    6) After migration finished, the VM goes to
                    VERIFY_RESIZE status, and "nova show" indicates that
                    the VM now located on host1:Cluster1<br>
                    | OS-EXT-SRV-ATTR:host                 | host1 |<br>
                    | OS-EXT-SRV-ATTR:hypervisor_hostname  |
                    domain-c12(Cluster1)                                
                        |<br>
                    7) But from vSphere client, it indicates the the VM
                    was still running on Cluster2<br>
                    8) Try to confirm the resize, confirm will be
                    failed. The root cause is that nova compute on host2
                    has no knowledge of domain-c12(Cluster1)<br>
                    <br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2810, in do_confirm_resize<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp    
                    migration=migration)<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2836, in _confirm_resize<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp     network_info)<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 420, in confirm_migration<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp     _vmops =
                    self._get_vmops_for_compute_node(instance['node'])<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 523, in _get_vmops_for_compute_node<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp     resource =
                    self._get_resource_for_node(nodename)<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 515, in _get_resource_for_node<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp     raise
                    exception.NotFound(msg)<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp NotFound: NV-3AB798A
                    The resource domain-c12(Cluster1) does not exist<br>
                    2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    <br>
                    <br>
                    Scenario 2:<br>
                    <br>
                    1) Started two nova computes manage the two
                    clusters, but the two computes have same nova conf.<br>
                    1) nova-compute1.conf<br>
                    cluster_name=Cluster1<br>
                    cluster_name=Cluster2<br>
                    <br>
                    2) nova-compute2.conf<br>
                    cluster_name=Cluster1<br>
                    cluster_name=Cluster2<br>
                    <br>
                    3) Then create and resize/cold migrate a VM, it can
                    always succeed.<br>
                    <br>
                    <br>
                    Questions:<br>
                    For multi-cluster management, does vmware require
                    all nova compute have same cluster configuration to
                    make sure resize/cold migration can succeed?<br>
                    <br>
                    --<br>
                    Thanks,<br>
                    <br>
                    Jay<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/0b71a846/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/0b71a846/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 2<br>
                    Date: Sun, 16 Feb 2014 22:52:01 +0800<br>
                    From: Dong Liu <<a href="mailto:willowd878@gmail.com" target="_blank">willowd878@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: [openstack-dev] [Neutron]Do you think
                    tanent_id should be<br>
                            verified<br>
                    Message-ID: <<a href="mailto:26565D39-5372-48A5-8299-34DDE6C3394D@gmail.com" target="_blank">26565D39-5372-48A5-8299-34DDE6C3394D@gmail.com</a>><br>
                    Content-Type: text/plain; charset=us-ascii<br>
                    <br>
                    Hi stackers:<br>
                    <br>
                    I found that when creating network subnet and other
                    resources, the attribute tenant_id<br>
                    can be set by admin tenant. But we did not verify
                    that if the tanent_id is real in keystone.<br>
                    <br>
                    I know that we could use neutron without keystone,
                    but do you think tenant_id should<br>
                    be verified when we using neutron with keystone.<br>
                    <br>
                    thanks<br>
                    <br>
                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 3<br>
                    Date: Sun, 16 Feb 2014 23:01:17 +0800<br>
                    From: Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to<br>
                            resize/cold migration<br>
                    Message-ID:<br>
                            <<a href="mailto:CAFyztAFqTUqTZzzW6BkH6-9_kye9ZGm8yhZe3hMUoW1xFfQM7A@mail.gmail.com" target="_blank">CAFyztAFqTUqTZzzW6BkH6-9_kye9ZGm8yhZe3hMUoW1xFfQM7A@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset="iso-8859-1"<br>
                    <br>
                    Thanks Gary, clear now. ;-)<br>
                    <br>
                    <br>
                    2014-02-16 21:40 GMT+08:00 Gary Kotton <<a href="mailto:gkotton@vmware.com" target="_blank">gkotton@vmware.com</a>>:<br>
                    <br>
                    > Hi,<br>
                    > There are two issues here.<br>
                    > The first is a bug fix that is in review:<br>
                    > - <a href="https://review.openstack.org/#/c/69209/" target="_blank">https://review.openstack.org/#/c/69209/</a>
                    (this is where they have the<br>
                    > same configuration)<br>
                    > The second is WIP:<br>
                    > - <a href="https://review.openstack.org/#/c/69262/" target="_blank">https://review.openstack.org/#/c/69262/</a>
                    (we need to restore)<br>
                    > Thanks<br>
                    > Gary<br>
                    ><br>
                    > From: Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>><br>
                    > Reply-To: "OpenStack Development Mailing List
                    (not for usage questions)" <<br>
                    > <a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    > Date: Sunday, February 16, 2014 6:39 AM<br>
                    > To: OpenStack Development Mailing List <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    > Subject: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to<br>
                    > resize/cold migration<br>
                    ><br>
                    > Hey,<br>
                    ><br>
                    > I have one question related with OpenStack
                    vmwareapi.VMwareVCDriver<br>
                    > resize/cold migration.<br>
                    ><br>
                    > The following is my configuration:<br>
                    ><br>
                    >  DC<br>
                    >     |<br>
                    >     |----Cluster1<br>
                    >     |          |<br>
                    >     |          |----9.111.249.56<br>
                    >     |<br>
                    >     |----Cluster2<br>
                    >                |<br>
                    >                |----9.111.249.49<br>
                    ><br>
                    > *Scenario 1:*<br>
                    > I started two nova computes manage the two
                    clusters:<br>
                    > 1) nova-compute1.conf<br>
                    > cluster_name=Cluster1<br>
                    ><br>
                    > 2) nova-compute2.conf<br>
                    > cluster_name=Cluster2<br>
                    ><br>
                    > 3) Start up two nova computes on host1 and
                    host2 separately<br>
                    > 4) Create one VM instance and the VM instance
                    was booted on Cluster2 node<br>
                    > 9.111.249.49<br>
                    > | OS-EXT-SRV-ATTR:host                 | host2
                    |<br>
                    > | OS-EXT-SRV-ATTR:hypervisor_hostname  |<br>
                    > domain-c16(Cluster2)                          
                              |<br>
                    > 5) Cold migrate the VM instance<br>
                    > 6) After migration finished, the VM goes to
                    VERIFY_RESIZE status, and<br>
                    > "nova show" indicates that the VM now located
                    on host1:Cluster1<br>
                    > | OS-EXT-SRV-ATTR:host                 | host1
                    |<br>
                    > | OS-EXT-SRV-ATTR:hypervisor_hostname  |<br>
                    > domain-c12(Cluster1)                          
                              |<br>
                    > 7) But from vSphere client, it indicates the
                    the VM was still running on<br>
                    > Cluster2<br>
                    > 8) Try to confirm the resize, confirm will be
                    failed. The root cause is<br>
                    > that nova compute on host2 has no knowledge of
                    domain-c12(Cluster1)<br>
                    ><br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2810, in<br>
                    > do_confirm_resize<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > migration=migration)<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2836, in<br>
                    > _confirm_resize<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > network_info)<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 420,<br>
                    > in confirm_migration<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > _vmops =
                    self._get_vmops_for_compute_node(instance['node'])<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 523,<br>
                    > in _get_vmops_for_compute_node<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > resource =
                    self._get_resource_for_node(nodename)<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 515,<br>
                    > in _get_resource_for_node<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > raise exception.NotFound(msg)<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    > NotFound: NV-3AB798A The resource
                    domain-c12(Cluster1) does not exist<br>
                    > 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    ><br>
                    ><br>
                    > *Scenario 2:*<br>
                    ><br>
                    > 1) Started two nova computes manage the two
                    clusters, but the two computes<br>
                    > have same nova conf.<br>
                    > 1) nova-compute1.conf<br>
                    > cluster_name=Cluster1<br>
                    > cluster_name=Cluster2<br>
                    ><br>
                    > 2) nova-compute2.conf<br>
                    > cluster_name=Cluster1<br>
                    > cluster_name=Cluster2<br>
                    ><br>
                    > 3) Then create and resize/cold migrate a VM, it
                    can always succeed.<br>
                    ><br>
                    ><br>
                    > *Questions:*<br>
                    > For multi-cluster management, does vmware
                    require all nova compute have<br>
                    > same cluster configuration to make sure
                    resize/cold migration can succeed?<br>
                    ><br>
                    > --<br>
                    > Thanks,<br>
                    ><br>
                    > Jay<br>
                    <div>><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      ><br>
                      <br>
                      <br>
                    </div>
                    --<br>
                    Thanks,<br>
                    <br>
                    Jay<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/a5a2ed40/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/a5a2ed40/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 4<br>
                    Date: Sun, 16 Feb 2014 10:27:41 -0500<br>
                    From: Mohammad Banikazemi <<a href="mailto:mb@us.ibm.com" target="_blank">mb@us.ibm.com</a>><br>
                    To: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    Subject: [openstack-dev] [neutron][policy] Using
                    network services with<br>
                            network policies<br>
                    Message-ID:<br>
                            <<a href="mailto:OF456914EA.334156E1-ON85257C81.0051DB09-85257C81.0054EF2C@us.ibm.com" target="_blank">OF456914EA.334156E1-ON85257C81.0051DB09-85257C81.0054EF2C@us.ibm.com</a>><br>
                    Content-Type: text/plain; charset="us-ascii"<br>
                    <br>
                    <br>
                    During the last IRC call we started talking about
                    network services and how<br>
                    they can be integrated into the group Policy
                    framework.<br>
                    <br>
                    In particular, with the "redirect" action we need to
                    think how we can<br>
                    specify the network services we want to redirect the
                    traffic to/from. There<br>
                    has been a substantial work in the area of service
                    chaining and service<br>
                    insertion and in the last summit "advanced service"
                    in VMs were discussed.<br>
                    I think the first step for us is to find out the
                    status of those efforts<br>
                    and then see how we can use them. Here are a few
                    questions that come to<br>
                    mind.<br>
                    1- What is the status of service chaining, service
                    insertion and advanced<br>
                    services work?<br>
                    2- How could we use a service chain? Would simply
                    referring to it in the<br>
                    action be enough? Are there considerations wrt
                    creating a service chain<br>
                    and/or a service VM for use with the Group Policy
                    framework that need to be<br>
                    taken into account?<br>
                    <br>
                    Let's start the discussion on the ML before taking
                    it to the next call.<br>
                    <br>
                    Thanks,<br>
                    <br>
                    Mohammad<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/efff7427/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/efff7427/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 5<br>
                    Date: Sun, 16 Feb 2014 23:29:49 +0800<br>
                    From: Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to<br>
                            resize/cold migration<br>
                    Message-ID:<br>
                            <<a href="mailto:CAFyztAFCc1NH5nz00Dii3dhL3AN8RjPLb3D65aFMRGfyQiJGKA@mail.gmail.com" target="_blank">CAFyztAFCc1NH5nz00Dii3dhL3AN8RjPLb3D65aFMRGfyQiJGKA@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset="iso-8859-1"<br>
                    <br>
                    Hi Gary,<br>
                    <br>
                    One more question, when using VCDriver, I can use it
                    in the following two<br>
                    ways:<br>
                    1) start up many nova computes and those nova
                    computes manage same vcenter<br>
                    clusters.<br>
                    2) start up many nova computes and those nova
                    computes manage different<br>
                    vcenter clusters.<br>
                    <br>
                    Do we have some best practice for above two
                    scenarios or else can you<br>
                    please provide some best practise for VCDriver? I
                    did not get much info<br>
                    from admin guide.<br>
                    <br>
                    Thanks,<br>
                    <br>
                    Jay<br>
                    <br>
                    <br>
                    2014-02-16 23:01 GMT+08:00 Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>>:<br>
                    <br>
                    > Thanks Gary, clear now. ;-)<br>
                    ><br>
                    ><br>
                    > 2014-02-16 21:40 GMT+08:00 Gary Kotton <<a href="mailto:gkotton@vmware.com" target="_blank">gkotton@vmware.com</a>>:<br>
                    ><br>
                    >> Hi,<br>
                    >> There are two issues here.<br>
                    >> The first is a bug fix that is in review:<br>
                    >> - <a href="https://review.openstack.org/#/c/69209/" target="_blank">https://review.openstack.org/#/c/69209/</a>
                    (this is where they have the<br>
                    >> same configuration)<br>
                    >> The second is WIP:<br>
                    >> - <a href="https://review.openstack.org/#/c/69262/" target="_blank">https://review.openstack.org/#/c/69262/</a>
                    (we need to restore)<br>
                    >> Thanks<br>
                    >> Gary<br>
                    >><br>
                    >> From: Jay Lau <<a href="mailto:jay.lau.513@gmail.com" target="_blank">jay.lau.513@gmail.com</a>><br>
                    >> Reply-To: "OpenStack Development Mailing
                    List (not for usage questions)"<br>
                    >> <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    >> Date: Sunday, February 16, 2014 6:39 AM<br>
                    >> To: OpenStack Development Mailing List <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a><br>
                    >> ><br>
                    >> Subject: [openstack-dev] [Nova][VMWare]
                    VMwareVCDriver related to<br>
                    >> resize/cold migration<br>
                    >><br>
                    >> Hey,<br>
                    >><br>
                    >> I have one question related with OpenStack
                    vmwareapi.VMwareVCDriver<br>
                    >> resize/cold migration.<br>
                    >><br>
                    >> The following is my configuration:<br>
                    >><br>
                    >>  DC<br>
                    >>     |<br>
                    >>     |----Cluster1<br>
                    >>     |          |<br>
                    >>     |          |----9.111.249.56<br>
                    >>     |<br>
                    >>     |----Cluster2<br>
                    >>                |<br>
                    >>                |----9.111.249.49<br>
                    >><br>
                    >> *Scenario 1:*<br>
                    >> I started two nova computes manage the two
                    clusters:<br>
                    >> 1) nova-compute1.conf<br>
                    >> cluster_name=Cluster1<br>
                    >><br>
                    >> 2) nova-compute2.conf<br>
                    >> cluster_name=Cluster2<br>
                    >><br>
                    >> 3) Start up two nova computes on host1 and
                    host2 separately<br>
                    >> 4) Create one VM instance and the VM
                    instance was booted on Cluster2<br>
                    >> node  9.111.249.49<br>
                    >> | OS-EXT-SRV-ATTR:host                 |
                    host2 |<br>
                    >> | OS-EXT-SRV-ATTR:hypervisor_hostname  |<br>
                    >> domain-c16(Cluster2)                      
                                  |<br>
                    >> 5) Cold migrate the VM instance<br>
                    >> 6) After migration finished, the VM goes to
                    VERIFY_RESIZE status, and<br>
                    >> "nova show" indicates that the VM now
                    located on host1:Cluster1<br>
                    >> | OS-EXT-SRV-ATTR:host                 |
                    host1 |<br>
                    >> | OS-EXT-SRV-ATTR:hypervisor_hostname  |<br>
                    >> domain-c12(Cluster1)                      
                                  |<br>
                    >> 7) But from vSphere client, it indicates
                    the the VM was still running on<br>
                    >> Cluster2<br>
                    >> 8) Try to confirm the resize, confirm will
                    be failed. The root cause is<br>
                    >> that nova compute on host2 has no knowledge
                    of domain-c12(Cluster1)<br>
                    >><br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >>
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2810, in<br>
                    >> do_confirm_resize<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> migration=migration)<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >>
                    "/usr/lib/python2.6/site-packages/nova/compute/manager.py",
                    line 2836, in<br>
                    >> _confirm_resize<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> network_info)<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >>
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 420,<br>
                    >> in confirm_migration<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> _vmops =
                    self._get_vmops_for_compute_node(instance['node'])<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >>
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 523,<br>
                    >> in _get_vmops_for_compute_node<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> resource =
                    self._get_resource_for_node(nodename)<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp   File<br>
                    >>
                    "/usr/lib/python2.6/site-packages/nova/virt/vmwareapi/driver.py",
                    line 515,<br>
                    >> in _get_resource_for_node<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> raise exception.NotFound(msg)<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >> NotFound: NV-3AB798A The resource
                    domain-c12(Cluster1) does not exist<br>
                    >> 2014-02-16 07:10:17.166 12720 TRACE
                    nova.openstack.common.rpc.amqp<br>
                    >><br>
                    >><br>
                    >> *Scenario 2:*<br>
                    >><br>
                    >> 1) Started two nova computes manage the two
                    clusters, but the two<br>
                    >> computes have same nova conf.<br>
                    >> 1) nova-compute1.conf<br>
                    >> cluster_name=Cluster1<br>
                    >> cluster_name=Cluster2<br>
                    >><br>
                    >> 2) nova-compute2.conf<br>
                    >> cluster_name=Cluster1<br>
                    >> cluster_name=Cluster2<br>
                    >><br>
                    >> 3) Then create and resize/cold migrate a
                    VM, it can always succeed.<br>
                    >><br>
                    >><br>
                    >> *Questions:*<br>
                    >> For multi-cluster management, does vmware
                    require all nova compute have<br>
                    >> same cluster configuration to make sure
                    resize/cold migration can succeed?<br>
                    >><br>
                    >> --<br>
                    >> Thanks,<br>
                    >><br>
                    >> Jay<br>
                    <div>>><br>
                      >>
                      _______________________________________________<br>
                      >> OpenStack-dev mailing list<br>
                      >> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >><br>
                      >><br>
                      ><br>
                      ><br>
                    </div>
                    > --<br>
                    > Thanks,<br>
                    ><br>
                    > Jay<br>
                    ><br>
                    <br>
                    <br>
                    <br>
                    --<br>
                    Thanks,<br>
                    <br>
                    Jay<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/e7da9e73/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/e7da9e73/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 6<br>
                    Date: Sun, 16 Feb 2014 08:01:14 -0800<br>
                    From: Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [keystone] role of
                    Domain in VPC<br>
                            definition<br>
                    Message-ID:
                    <-4426752061342328447@unknownmsgid><br>
                    Content-Type: text/plain; charset="iso-8859-1"<br>
                    <br>
                    Yes, [1] can be done without [2] and [3].<br>
                    As you are well aware [2] is now merged with group
                    policy discussions.<br>
                    IMHO all or nothing approach will not get us
                    anywhere.<br>
                    By the time we line up all our ducks in row. New
                    features/ideas/blueprints<br>
                    will keep Emerging.<br>
                    <br>
                    Regards<br>
                    -Harshad<br>
                    <br>
                    <br>
                    On Feb 16, 2014, at 2:30 AM, Salvatore Orlando <<a href="mailto:sorlando@nicira.com" target="_blank">sorlando@nicira.com</a>>

                    wrote:<br>
                    <br>
                    It seems this work item is made of several
                    blueprints, some of which are<br>
                    not yet approved. This is true at least for the
                    Neutron blueprint regarding<br>
                    policy extensions.<br>
                    <br>
                    Since I first looked at this spec I've been
                    wondering why nova has been<br>
                    selected as an endpoint for network operations
                    rather than Neutron, but<br>
                    this probably a design/implementation details
                    whereas JC here is looking at<br>
                    the general approach.<br>
                    <br>
                    Nevertheless, my only point here is that is seems
                    that features like this<br>
                    need an "all-or-none" approval.<br>
                    For instance, could the VPC feature be considered
                    functional if blueprint<br>
                    [1] is implemented, but not [2] and [3]?<br>
                    <br>
                    Salvatore<br>
                    <br>
                    [1] <a href="https://blueprints.launchpad.net/nova/+spec/aws-vpc-support" target="_blank">https://blueprints.launchpad.net/nova/+spec/aws-vpc-support</a><br>
                    [2]<br>
                    <a href="https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron" target="_blank">https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron</a><br>
                    [3]<br>
                    <a href="https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy" target="_blank">https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy</a><br>
                    <br>
                    <br>
                    On 11 February 2014 21:45, Martin, JC <<a href="mailto:jch.martin@gmail.com" target="_blank">jch.martin@gmail.com</a>>

                    wrote:<br>
                    <br>
                    > Ravi,<br>
                    ><br>
                    > It seems that the following Blueprint<br>
                    > <a href="https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support" target="_blank">https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support</a><br>
                    ><br>
                    > has been approved.<br>
                    ><br>
                    > However, I cannot find a discussion with regard
                    to the merit of using<br>
                    > project vs. domain, or other mechanism for the
                    implementation.<br>
                    ><br>
                    > I have an issue with this approach as it
                    prevents tenants within the same<br>
                    > domain sharing the same VPC to have projects.<br>
                    ><br>
                    > As an example, if you are a large organization
                    on AWS, it is likely that<br>
                    > you have a large VPC that will be shred by
                    multiple projects. With this<br>
                    > proposal, we loose that capability, unless I
                    missed something.<br>
                    ><br>
                    > JC<br>
                    ><br>
                    > On Dec 19, 2013, at 6:10 PM, Ravi Chunduru <<a href="mailto:ravivsn@gmail.com" target="_blank">ravivsn@gmail.com</a>>

                    wrote:<br>
                    ><br>
                    > > Hi,<br>
                    > >   We had some internal discussions on role
                    of Domain and VPCs. I would<br>
                    > like to expand and understand community
                    thinking of Keystone domain and<br>
                    > VPCs.<br>
                    > ><br>
                    > > Is VPC equivalent to Keystone Domain?<br>
                    > ><br>
                    > > If so, as a public cloud provider - I
                    create a Keystone domain and give<br>
                    > it to an organization which wants a virtual
                    private cloud.<br>
                    > ><br>
                    > > Now the question is if that organization
                    wants to have  departments wise<br>
                    > allocation of resources it is becoming
                    difficult to visualize with existing<br>
                    > v3 keystone constructs.<br>
                    > ><br>
                    > > Currently, it looks like each department
                    of an organization cannot have<br>
                    > their own resource management with in the
                    organization VPC ( LDAP based<br>
                    > user management, network management or
                    dedicating computes etc.,) For us,<br>
                    > Openstack Project does not match the
                    requirements of a department of an<br>
                    > organization.<br>
                    > ><br>
                    > > I hope you guessed what we wanted - Domain
                    must have VPCs and VPC to<br>
                    > have projects.<br>
                    > ><br>
                    > > I would like to know how community see the
                    VPC model in Openstack.<br>
                    > ><br>
                    > > Thanks,<br>
                    > > -Ravi.<br>
                    <div>> ><br>
                      > ><br>
                      > >
                      _______________________________________________<br>
                      > > OpenStack-dev mailing list<br>
                      > > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      <br>
                      _______________________________________________<br>
                      OpenStack-dev mailing list<br>
                      <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                    </div>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/9258cf27/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/9258cf27/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 7<br>
                    Date: Sun, 16 Feb 2014 08:47:19 -0800<br>
                    From: Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] VPC Proposal<br>
                    Message-ID:<br>
                            <<a href="mailto:CAL7PBMchfaSkX8amUAEe8X_fs9OM6ZLGJx_fNB2SUCJWPaGNFA@mail.gmail.com" target="_blank">CAL7PBMchfaSkX8amUAEe8X_fs9OM6ZLGJx_fNB2SUCJWPaGNFA@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset="iso-8859-1"<br>
                    <br>
                    Comments Inline<br>
                    <br>
                    Regards<br>
                    -Harshad<br>
                    <br>
                    <br>
                    On Sat, Feb 15, 2014 at 11:39 PM, Allamaraju, Subbu
                    <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>

                    wrote:<br>
                    <br>
                    > Harshad,<br>
                    ><br>
                    > Curious to know if there is a broad interest in
                    an AWS compatible API in<br>
                    > the community?<br>
                    <br>
                    <br>
                    We started looking at this as some our
                    customers/partners were interested<br>
                    in get AWS API compatibility. We have this blueprint
                    and code review<br>
                    pending for long time now. We will know based on
                    this thread wether the<br>
                    community is interested. But I assumed that
                    community was interested as the<br>
                    blueprint was approved and code review has no -1(s)
                    for long time now.<br>
                    <br>
                    <br>
                    > To clarify, a clear incremental path from an
                    AWS compatible API to an<br>
                    > OpenStack model is not clear.<br>
                    ><br>
                    <br>
                    In my mind AWS compatible API does not need new
                    openstack model. As more<br>
                    discussion happen on JC's proposal and
                    implementation becomes clear we will<br>
                    know how incremental is the path. But at high level
                    there two major<br>
                    differences<br>
                    1. New first class object will be introduced which
                    effect all components<br>
                    2. more than one project can be supported within
                    VPC.<br>
                    But it does not change AWS API(s). So even in JC(s)
                    model if you want AWS<br>
                    API then we will have to keep VPC to project mapping
                    1:1, since the API<br>
                    will not take both VPC ID and project ID.<br>
                    <br>
                    As more users want to migrate from AWS or IaaS
                    providers who want compete<br>
                    with AWS should be interested in this compatibility.<br>
                    <br>
                    There also seems to be terminology issue here Whats
                    is definition of "VPC"<br>
                    if we assume what AWS implements is "VPC"<br>
                    then what JC is proposing "VOS" or "VDC" (virtual
                    openstack or virtual DC)<br>
                    as all or most of current openstack features are
                    available to user in  this<br>
                    new Abstraction. I actually like this new
                    abstraction.<br>
                    <br>
                    <br>
                    > Subbu<br>
                    ><br>
                    > On Feb 15, 2014, at 10:04 PM, Harshad Nakil
                    <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    > wrote:<br>
                    ><br>
                    > ><br>
                    > > I agree with problem as defined by you and
                    will require more fundamental<br>
                    > changes.<br>
                    > > Meanwhile many users will benefit from AWS
                    VPC api compatibility.<br>
                    <div>><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                    </div>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/5f655f01/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/5f655f01/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 8<br>
                    Date: Sun, 16 Feb 2014 09:04:36 -0800<br>
                    From: "Allamaraju, Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>><br>
                    To: Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    Cc: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] VPC Proposal<br>
                    Message-ID: <<a href="mailto:641D4BA6-DFB2-4D3E-8D67-48F711ADC1B5@subbu.org" target="_blank">641D4BA6-DFB2-4D3E-8D67-48F711ADC1B5@subbu.org</a>><br>
                    Content-Type: text/plain; charset=iso-8859-1<br>
                    <br>
                    Harshad,<br>
                    <br>
                    Thanks for clarifying.<br>
                    <br>
                    > We started looking at this as some our
                    customers/partners were interested in get AWS API
                    compatibility. We have this blueprint and code
                    review pending for long time now. We will know based
                    on this thread wether the community is interested.
                    But I assumed that community was interested as the
                    blueprint was approved and code review has no -1(s)
                    for long time now.<br>
                    <br>
                    Makes sense. I would leave it to others on this list
                    to chime in if there is sufficient interest or not.<br>
                    <br>
                    > To clarify, a clear incremental path from an
                    AWS compatible API to an OpenStack model is not
                    clear.<br>
                    ><br>
                    > In my mind AWS compatible API does not need new
                    openstack model. As more discussion happen on JC's
                    proposal and implementation becomes clear we will
                    know how incremental is the path. But at high level
                    there two major differences<br>
                    > 1. New first class object will be introduced
                    which effect all components<br>
                    > 2. more than one project can be supported
                    within VPC.<br>
                    > But it does not change AWS API(s). So even in
                    JC(s) model if you want AWS API then we will have to
                    keep VPC to project mapping 1:1, since the API will
                    not take both VPC ID and project ID.<br>
                    ><br>
                    > As more users want to migrate from AWS or IaaS
                    providers who want compete with AWS should be
                    interested in this compatibility.<br>
                    <br>
                    IMHO that's a tough sell. Though an AWS compatible
                    API does not need an OpenStack abstraction, we would
                    end up with two independent ways of doing similar
                    things. That would OpenStack repeating itself!<br>
                    <br>
                    Subbu<br>
                    <br>
                    <br>
                    <br>
                    <br>
                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 9<br>
                    Date: Sun, 16 Feb 2014 09:12:54 -0800<br>
                    From: Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    To: "Allamaraju, Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>><br>
                    Cc: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] VPC Proposal<br>
                    Message-ID: <516707826958554641@unknownmsgid><br>
                    Content-Type: text/plain; charset=ISO-8859-1<br>
                    <br>
                    IMHO I don't see two implementations. Since right
                    now we have only<br>
                    one. As a community if we decide to add new
                    abstractions then we will<br>
                    have to change software in every component where the
                    new abstraction<br>
                    makes difference. That's normal software development
                    process.<br>
                    Regards<br>
                    -Harshad<br>
                    <br>
                    <br>
                    > On Feb 16, 2014, at 9:03 AM, "Allamaraju,
                    Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>

                    wrote:<br>
                    ><br>
                    > Harshad,<br>
                    ><br>
                    > Thanks for clarifying.<br>
                    ><br>
                    >> We started looking at this as some our
                    customers/partners were interested in get AWS API
                    compatibility. We have this blueprint and code
                    review pending for long time now. We will know based
                    on this thread wether the community is interested.
                    But I assumed that community was interested as the
                    blueprint was approved and code review has no -1(s)
                    for long time now.<br>
                    ><br>
                    > Makes sense. I would leave it to others on this
                    list to chime in if there is sufficient interest or
                    not.<br>
                    ><br>
                    >> To clarify, a clear incremental path from
                    an AWS compatible API to an OpenStack model is not
                    clear.<br>
                    >><br>
                    >> In my mind AWS compatible API does not need
                    new openstack model. As more discussion happen on
                    JC's proposal and implementation becomes clear we
                    will know how incremental is the path. But at high
                    level there two major differences<br>
                    >> 1. New first class object will be
                    introduced which effect all components<br>
                    >> 2. more than one project can be supported
                    within VPC.<br>
                    >> But it does not change AWS API(s). So even
                    in JC(s) model if you want AWS API then we will have
                    to keep VPC to project mapping 1:1, since the API
                    will not take both VPC ID and project ID.<br>
                    >><br>
                    >> As more users want to migrate from AWS or
                    IaaS providers who want compete with AWS should be
                    interested in this compatibility.<br>
                    ><br>
                    > IMHO that's a tough sell. Though an AWS
                    compatible API does not need an OpenStack
                    abstraction, we would end up with two independent
                    ways of doing similar things. That would OpenStack
                    repeating itself!<br>
                    ><br>
                    > Subbu<br>
                    ><br>
                    ><br>
                    <br>
                    <br>
                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 10<br>
                    Date: Sun, 16 Feb 2014 09:25:02 -0800<br>
                    From: "Martin, JC" <<a href="mailto:jch.martin@gmail.com" target="_blank">jch.martin@gmail.com</a>><br>
                    To: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] VPC Proposal<br>
                    Message-ID: <<a href="mailto:B1A58385-DC10-48EF-AA8E-90176F576A40@gmail.com" target="_blank">B1A58385-DC10-48EF-AA8E-90176F576A40@gmail.com</a>><br>
                    Content-Type: text/plain; charset=us-ascii<br>
                    <br>
                    Harshad,<br>
                    <br>
                    I tried to find some discussion around this
                    blueprint.<br>
                    Could you provide us with some notes or threads  ?<br>
                    <br>
                    Also, about the code review you mention. which one
                    are you talking about :<br>
                    <a href="https://review.openstack.org/#/c/40071/" target="_blank">https://review.openstack.org/#/c/40071/</a><br>
                    <a href="https://review.openstack.org/#/c/49470/" target="_blank">https://review.openstack.org/#/c/49470/</a><br>
                    <a href="https://review.openstack.org/#/c/53171" target="_blank">https://review.openstack.org/#/c/53171</a><br>
                    <br>
                    because they are all abandoned.<br>
                    <br>
                    Could you point me to the code, and update the BP
                    because it seems that the links are not correct.<br>
                    <br>
                    Thanks,<br>
                    <br>
                    JC<br>
                    On Feb 16, 2014, at 9:04 AM, "Allamaraju, Subbu"
                    <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>

                    wrote:<br>
                    <br>
                    > Harshad,<br>
                    ><br>
                    > Thanks for clarifying.<br>
                    ><br>
                    >> We started looking at this as some our
                    customers/partners were interested in get AWS API
                    compatibility. We have this blueprint and code
                    review pending for long time now. We will know based
                    on this thread wether the community is interested.
                    But I assumed that community was interested as the
                    blueprint was approved and code review has no -1(s)
                    for long time now.<br>
                    ><br>
                    > Makes sense. I would leave it to others on this
                    list to chime in if there is sufficient interest or
                    not.<br>
                    ><br>
                    >> To clarify, a clear incremental path from
                    an AWS compatible API to an OpenStack model is not
                    clear.<br>
                    >><br>
                    >> In my mind AWS compatible API does not need
                    new openstack model. As more discussion happen on
                    JC's proposal and implementation becomes clear we
                    will know how incremental is the path. But at high
                    level there two major differences<br>
                    >> 1. New first class object will be
                    introduced which effect all components<br>
                    >> 2. more than one project can be supported
                    within VPC.<br>
                    >> But it does not change AWS API(s). So even
                    in JC(s) model if you want AWS API then we will have
                    to keep VPC to project mapping 1:1, since the API
                    will not take both VPC ID and project ID.<br>
                    >><br>
                    >> As more users want to migrate from AWS or
                    IaaS providers who want compete with AWS should be
                    interested in this compatibility.<br>
                    ><br>
                    > IMHO that's a tough sell. Though an AWS
                    compatible API does not need an OpenStack
                    abstraction, we would end up with two independent
                    ways of doing similar things. That would OpenStack
                    repeating itself!<br>
                    ><br>
                    > Subbu<br>
                    <div>><br>
                      ><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      <br>
                      <br>
                      <br>
                      <br>
                    </div>
                    ------------------------------<br>
                    <br>
                    Message: 11<br>
                    Date: Sun, 16 Feb 2014 09:49:17 -0800<br>
                    From: "Allamaraju, Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [keystone] role of
                    Domain in VPC<br>
                            definition<br>
                    Message-ID: <<a href="mailto:1756EFC4-ABAF-4377-B44A-219F34C3ABFA@subbu.org" target="_blank">1756EFC4-ABAF-4377-B44A-219F34C3ABFA@subbu.org</a>><br>
                    Content-Type: text/plain; charset=iso-8859-1<br>
                    <br>
                    Harshad,<br>
                    <br>
                    But the key question that Ravi brought up remains
                    though. A project is a very small administrative
                    container to manage policies and resources for VPCs.
                    We've been experimenting with VPCs on OpenStack
                    (with some mods) at work for nearly a year, and came
                    across cases where hundreds/thousands of apps in
                    equal number of projects needing to share resources
                    and policies, and project to VPC mapping did not
                    cut.<br>
                    <br>
                    I was wondering if there was prior discussion around
                    the mapping of AWS VPC model to OpenStack concepts
                    like projects and domains. Thanks for any pointers.<br>
                    <br>
                    Subbu<br>
                    <br>
                    On Feb 16, 2014, at 8:01 AM, Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>>

                    wrote:<br>
                    <br>
                    > Yes, [1] can be done without [2] and [3].<br>
                    > As you are well aware [2] is now merged with
                    group policy discussions.<br>
                    > IMHO all or nothing approach will not get us
                    anywhere.<br>
                    > By the time we line up all our ducks in row.
                    New features/ideas/blueprints will keep Emerging.<br>
                    ><br>
                    > Regards<br>
                    > -Harshad<br>
                    ><br>
                    ><br>
                    > On Feb 16, 2014, at 2:30 AM, Salvatore Orlando
                    <<a href="mailto:sorlando@nicira.com" target="_blank">sorlando@nicira.com</a>>

                    wrote:<br>
                    ><br>
                    >> It seems this work item is made of several
                    blueprints, some of which are not yet approved. This
                    is true at least for the Neutron blueprint regarding
                    policy extensions.<br>
                    >><br>
                    >> Since I first looked at this spec I've been
                    wondering why nova has been selected as an endpoint
                    for network operations rather than Neutron, but this
                    probably a design/implementation details whereas JC
                    here is looking at the general approach.<br>
                    >><br>
                    >> Nevertheless, my only point here is that is
                    seems that features like this need an "all-or-none"
                    approval.<br>
                    >> For instance, could the VPC feature be
                    considered functional if blueprint [1] is
                    implemented, but not [2] and [3]?<br>
                    >><br>
                    >> Salvatore<br>
                    >><br>
                    >> [1] <a href="https://blueprints.launchpad.net/nova/+spec/aws-vpc-support" target="_blank">https://blueprints.launchpad.net/nova/+spec/aws-vpc-support</a><br>
                    >> [2] <a href="https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron" target="_blank">https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron</a><br>


                    >> [3] <a href="https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy" target="_blank">https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy</a><br>
                    >><br>
                    >><br>
                    >> On 11 February 2014 21:45, Martin, JC <<a href="mailto:jch.martin@gmail.com" target="_blank">jch.martin@gmail.com</a>>

                    wrote:<br>
                    >> Ravi,<br>
                    >><br>
                    >> It seems that the following Blueprint<br>
                    >> <a href="https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support" target="_blank">https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support</a><br>
                    >><br>
                    >> has been approved.<br>
                    >><br>
                    >> However, I cannot find a discussion with
                    regard to the merit of using project vs. domain, or
                    other mechanism for the implementation.<br>
                    >><br>
                    >> I have an issue with this approach as it
                    prevents tenants within the same domain sharing the
                    same VPC to have projects.<br>
                    >><br>
                    >> As an example, if you are a large
                    organization on AWS, it is likely that you have a
                    large VPC that will be shred by multiple projects.
                    With this proposal, we loose that capability, unless
                    I missed something.<br>
                    >><br>
                    >> JC<br>
                    >><br>
                    >> On Dec 19, 2013, at 6:10 PM, Ravi Chunduru
                    <<a href="mailto:ravivsn@gmail.com" target="_blank">ravivsn@gmail.com</a>>

                    wrote:<br>
                    >><br>
                    >> > Hi,<br>
                    >> >   We had some internal discussions on
                    role of Domain and VPCs. I would like to expand and
                    understand community thinking of Keystone domain and
                    VPCs.<br>
                    >> ><br>
                    >> > Is VPC equivalent to Keystone Domain?<br>
                    >> ><br>
                    >> > If so, as a public cloud provider - I
                    create a Keystone domain and give it to an
                    organization which wants a virtual private cloud.<br>
                    >> ><br>
                    >> > Now the question is if that
                    organization wants to have  departments wise
                    allocation of resources it is becoming difficult to
                    visualize with existing v3 keystone constructs.<br>
                    >> ><br>
                    >> > Currently, it looks like each
                    department of an organization cannot have their own
                    resource management with in the organization VPC (
                    LDAP based user management, network management or
                    dedicating computes etc.,) For us, Openstack Project
                    does not match the requirements of a department of
                    an organization.<br>
                    >> ><br>
                    >> > I hope you guessed what we wanted -
                    Domain must have VPCs and VPC to have projects.<br>
                    >> ><br>
                    >> > I would like to know how community see
                    the VPC model in Openstack.<br>
                    >> ><br>
                    >> > Thanks,<br>
                    >> > -Ravi.<br>
                    <div>>> ><br>
                      >> ><br>
                      >> >
                      _______________________________________________<br>
                      >> > OpenStack-dev mailing list<br>
                      >> > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >><br>
                      >><br>
                      >>
                      _______________________________________________<br>
                      >> OpenStack-dev mailing list<br>
                      >> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >><br>
                      >>
                      _______________________________________________<br>
                      >> OpenStack-dev mailing list<br>
                      >> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      <br>
                      <br>
                      <br>
                      <br>
                    </div>
                    ------------------------------<br>
                    <br>
                    Message: 12<br>
                    Date: Sun, 16 Feb 2014 10:15:11 -0800<br>
                    From: Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [keystone] role of
                    Domain in VPC<br>
                            definition<br>
                    Message-ID: <4920517322402852354@unknownmsgid><br>
                    Content-Type: text/plain; charset=ISO-8859-1<br>
                    <br>
                    As said I am not disagreeing with you or Ravi or JC.
                    I also agree that<br>
                    Openstack VPC implementation will benefit from these
                    proposals.<br>
                    What I am saying is it is not required AWS VPC API
                    compatibility at<br>
                    this point.  Which is what our blueprint is all
                    about. We are not<br>
                    defining THE "VPC".<br>
                    Let me ask you what changes in AWS API when you go
                    to other model?<br>
                    The argument is you want multiple projects in VPC.
                    That's great. But I<br>
                    don't understand how I would specify it if my code
                    was written to use<br>
                    AWS API.<br>
                    The argument you want multiple external networks per
                    VPC I don't know<br>
                    how to specify using AWS API<br>
                    So list goes on.<br>
                    <br>
                    May be I am missing something. If you don't want AWS
                    compatibility<br>
                    then that's different issue all together. And should
                    be discussed as<br>
                    such.<br>
                    <br>
                    Regards<br>
                    -Harshad<br>
                    <br>
                    <br>
                    > On Feb 16, 2014, at 9:51 AM, "Allamaraju,
                    Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>

                    wrote:<br>
                    ><br>
                    > Harshad,<br>
                    ><br>
                    > But the key question that Ravi brought up
                    remains though. A project is a very small
                    administrative container to manage policies and
                    resources for VPCs. We've been experimenting with
                    VPCs on OpenStack (with some mods) at work for
                    nearly a year, and came across cases where
                    hundreds/thousands of apps in equal number of
                    projects needing to share resources and policies,
                    and project to VPC mapping did not cut.<br>
                    ><br>
                    > I was wondering if there was prior discussion
                    around the mapping of AWS VPC model to OpenStack
                    concepts like projects and domains. Thanks for any
                    pointers.<br>
                    ><br>
                    > Subbu<br>
                    ><br>
                    >> On Feb 16, 2014, at 8:01 AM, Harshad Nakil
                    <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>>

                    wrote:<br>
                    >><br>
                    >> Yes, [1] can be done without [2] and [3].<br>
                    >> As you are well aware [2] is now merged
                    with group policy discussions.<br>
                    >> IMHO all or nothing approach will not get
                    us anywhere.<br>
                    >> By the time we line up all our ducks in
                    row. New features/ideas/blueprints will keep
                    Emerging.<br>
                    >><br>
                    >> Regards<br>
                    >> -Harshad<br>
                    >><br>
                    >><br>
                    >>> On Feb 16, 2014, at 2:30 AM, Salvatore
                    Orlando <<a href="mailto:sorlando@nicira.com" target="_blank">sorlando@nicira.com</a>>

                    wrote:<br>
                    >>><br>
                    >>> It seems this work item is made of
                    several blueprints, some of which are not yet
                    approved. This is true at least for the Neutron
                    blueprint regarding policy extensions.<br>
                    >>><br>
                    >>> Since I first looked at this spec I've
                    been wondering why nova has been selected as an
                    endpoint for network operations rather than Neutron,
                    but this probably a design/implementation details
                    whereas JC here is looking at the general approach.<br>
                    >>><br>
                    >>> Nevertheless, my only point here is
                    that is seems that features like this need an
                    "all-or-none" approval.<br>
                    >>> For instance, could the VPC feature be
                    considered functional if blueprint [1] is
                    implemented, but not [2] and [3]?<br>
                    >>><br>
                    >>> Salvatore<br>
                    >>><br>
                    >>> [1] <a href="https://blueprints.launchpad.net/nova/+spec/aws-vpc-support" target="_blank">https://blueprints.launchpad.net/nova/+spec/aws-vpc-support</a><br>
                    >>> [2] <a href="https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron" target="_blank">https://blueprints.launchpad.net/neutron/+spec/policy-extensions-for-neutron</a><br>


                    >>> [3] <a href="https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy" target="_blank">https://blueprints.launchpad.net/keystone/+spec/hierarchical-multitenancy</a><br>
                    >>><br>
                    >>><br>
                    >>> On 11 February 2014 21:45, Martin, JC
                    <<a href="mailto:jch.martin@gmail.com" target="_blank">jch.martin@gmail.com</a>>

                    wrote:<br>
                    >>> Ravi,<br>
                    >>><br>
                    >>> It seems that the following Blueprint<br>
                    >>> <a href="https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support" target="_blank">https://wiki.openstack.org/wiki/Blueprint-aws-vpc-support</a><br>
                    >>><br>
                    >>> has been approved.<br>
                    >>><br>
                    >>> However, I cannot find a discussion
                    with regard to the merit of using project vs.
                    domain, or other mechanism for the implementation.<br>
                    >>><br>
                    >>> I have an issue with this approach as
                    it prevents tenants within the same domain sharing
                    the same VPC to have projects.<br>
                    >>><br>
                    >>> As an example, if you are a large
                    organization on AWS, it is likely that you have a
                    large VPC that will be shred by multiple projects.
                    With this proposal, we loose that capability, unless
                    I missed something.<br>
                    >>><br>
                    >>> JC<br>
                    >>><br>
                    >>>> On Dec 19, 2013, at 6:10 PM, Ravi
                    Chunduru <<a href="mailto:ravivsn@gmail.com" target="_blank">ravivsn@gmail.com</a>>

                    wrote:<br>
                    >>>><br>
                    >>>> Hi,<br>
                    >>>>  We had some internal discussions
                    on role of Domain and VPCs. I would like to expand
                    and understand community thinking of Keystone domain
                    and VPCs.<br>
                    >>>><br>
                    >>>> Is VPC equivalent to Keystone
                    Domain?<br>
                    >>>><br>
                    >>>> If so, as a public cloud provider -
                    I create a Keystone domain and give it to an
                    organization which wants a virtual private cloud.<br>
                    >>>><br>
                    >>>> Now the question is if that
                    organization wants to have  departments wise
                    allocation of resources it is becoming difficult to
                    visualize with existing v3 keystone constructs.<br>
                    >>>><br>
                    >>>> Currently, it looks like each
                    department of an organization cannot have their own
                    resource management with in the organization VPC (
                    LDAP based user management, network management or
                    dedicating computes etc.,) For us, Openstack Project
                    does not match the requirements of a department of
                    an organization.<br>
                    >>>><br>
                    >>>> I hope you guessed what we wanted -
                    Domain must have VPCs and VPC to have projects.<br>
                    >>>><br>
                    >>>> I would like to know how community
                    see the VPC model in Openstack.<br>
                    >>>><br>
                    >>>> Thanks,<br>
                    >>>> -Ravi.<br>
                    <div>>>>><br>
                      >>>><br>
                      >>>>
                      _______________________________________________<br>
                      >>>> OpenStack-dev mailing list<br>
                      >>>> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >>>> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >>><br>
                      >>><br>
                      >>>
                      _______________________________________________<br>
                      >>> OpenStack-dev mailing list<br>
                      >>> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >>> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >>><br>
                      >>>
                      _______________________________________________<br>
                      >>> OpenStack-dev mailing list<br>
                      >>> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >>> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >>
                      _______________________________________________<br>
                      >> OpenStack-dev mailing list<br>
                      >> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      <br>
                      <br>
                      <br>
                    </div>
                    ------------------------------<br>
                    <br>
                    Message: 13<br>
                    Date: Sun, 16 Feb 2014 10:31:42 -0800<br>
                    From: "Allamaraju, Subbu" <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [keystone] role of
                    Domain in VPC<br>
                            definition<br>
                    Message-ID: <<a href="mailto:7CD9E46E-FC0A-431B-836F-9BD02B0E417A@subbu.org" target="_blank">7CD9E46E-FC0A-431B-836F-9BD02B0E417A@subbu.org</a>><br>
                    Content-Type: text/plain; charset=us-ascii<br>
                    <br>
                    Harshad,<br>
                    <br>
                    This is great. At least there is consensus on what
                    it is and what it is not. I would leave it to others
                    to discuss merits of a an AWS compat VPC API for
                    Icehouse.<br>
                    <br>
                    Perhaps this is a good topic to discuss at the Juno
                    design summit.<br>
                    <br>
                    Subbu<br>
                    <br>
                    On Feb 16, 2014, at 10:15 AM, Harshad Nakil <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>>

                    wrote:<br>
                    <br>
                    > As said I am not disagreeing with you or Ravi
                    or JC. I also agree that<br>
                    > Openstack VPC implementation will benefit from
                    these proposals.<br>
                    > What I am saying is it is not required AWS VPC
                    API compatibility at<br>
                    > this point.  Which is what our blueprint is all
                    about. We are not<br>
                    > defining THE "VPC".<br>
                    <br>
                    <br>
                    <br>
                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 14<br>
                    Date: Mon, 17 Feb 2014 08:20:09 +1300<br>
                    From: Robert Collins <<a href="mailto:robertc@robertcollins.net" target="_blank">robertc@robertcollins.net</a>><br>
                    To: Sean Dague <<a href="mailto:sean@dague.net" target="_blank">sean@dague.net</a>><br>
                    Cc: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>>,<br>
                            "<<a href="mailto:openstack-infra@lists.openstack.org" target="_blank">openstack-infra@lists.openstack.org</a>>"<br>
                            <<a href="mailto:openstack-infra@lists.openstack.org" target="_blank">openstack-infra@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] [OpenStack-Infra]
                    [TripleO] promoting<br>
                            devtest_seed and devtest_undercloud to
                    voting, + experimental queue<br>
                            for nova/neutron etc.<br>
                    Message-ID:<br>
                           
                    <CAJ3HoZ1LC1WqayW3o3RaPxfLC0G-Lb9zxHKftPDW=<a href="mailto:t8wnubCtQ@mail.gmail.com" target="_blank">t8wnubCtQ@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset=ISO-8859-1<br>
                    <br>
                    On 15 February 2014 09:58, Sean Dague <<a href="mailto:sean@dague.net" target="_blank">sean@dague.net</a>>

                    wrote:<br>
                    <br>
                    >> Lastly, I'm going to propose a merge to
                    infra/config to put our<br>
                    >> undercloud story (which exercises the
                    seed's ability to deploy via<br>
                    >> heat with bare metal) as a check
                    experimental job on our dependencies<br>
                    >> (keystone, glance, nova, neutron) - if
                    thats ok with those projects?<br>
                    >><br>
                    >> -Rob<br>
                    >><br>
                    ><br>
                    > My biggest concern with adding this to check
                    experimental, is the<br>
                    > experimental results aren't published back
                    until all the experimental<br>
                    > jobs are done.<br>
                    <br>
                    If we add a new pipeline - <a href="https://review.openstack.org/#/c/73863/" target="_blank">https://review.openstack.org/#/c/73863/</a>
                    -<br>
                    then we can avoid that.<br>
                    <br>
                    > We've seen really substantial delays, plus a 5
                    day complete outage a<br>
                    > week ago, on the tripleo cloud. I'd like to see
                    that much more proven<br>
                    > before it starts to impact core projects, even
                    in experimental.<br>
                    <br>
                    I believe that with a new pipeline it won't impact
                    core projects at all.<br>
                    <br>
                    The outage, FWIW, was because I deleted the entire
                    cloud, at the same<br>
                    time that we had a firedrill with some other
                    upstream-of-us issue (I<br>
                    forget the exact one). The multi-region setup we're
                    aiming for should<br>
                    mitigate that substantially :)<br>
                    <div><br>
                      <br>
                      -Rob<br>
                      <br>
                      <br>
                      --<br>
                      Robert Collins <<a href="mailto:rbtcollins@hp.com" target="_blank">rbtcollins@hp.com</a>><br>
                      Distinguished Technologist<br>
                      HP Converged Cloud<br>
                      <br>
                      <br>
                      <br>
                      ------------------------------<br>
                      <br>
                    </div>
                    Message: 15<br>
                    Date: Mon, 17 Feb 2014 08:25:04 +1300<br>
                    From: Robert Collins <<a href="mailto:robertc@robertcollins.net" target="_blank">robertc@robertcollins.net</a>><br>
                    To: "James E. Blair" <<a href="mailto:jeblair@openstack.org" target="_blank">jeblair@openstack.org</a>><br>
                    Cc: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>>,<br>
                            "<<a href="mailto:openstack-infra@lists.openstack.org" target="_blank">openstack-infra@lists.openstack.org</a>>"<br>
                            <<a href="mailto:openstack-infra@lists.openstack.org" target="_blank">openstack-infra@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] [OpenStack-Infra]
                    [TripleO] promoting<br>
                            devtest_seed and devtest_undercloud to
                    voting, + experimental queue<br>
                            for nova/neutron etc.<br>
                    Message-ID:<br>
                            <<a href="mailto:CAJ3HoZ0me0xfeGArVSqLkC0SPpJwaTeK%2BhNYoePDdh_2FR_K9w@mail.gmail.com" target="_blank">CAJ3HoZ0me0xfeGArVSqLkC0SPpJwaTeK+hNYoePDdh_2FR_K9w@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset=ISO-8859-1<br>
                    <br>
                    On 15 February 2014 12:21, James E. Blair <<a href="mailto:jeblair@openstack.org" target="_blank">jeblair@openstack.org</a>>

                    wrote:<br>
                    <br>
                    > You won't end up with -1's everywhere, you'll
                    end up with jobs stuck in<br>
                    > the queue indefinitely, as we saw when the
                    tripleo cloud failed<br>
                    > recently.  What's worse is that now that
                    positive check results are<br>
                    > required for enqueuing into the gate, you will
                    also not be able to merge<br>
                    > anything.<br>
                    <br>
                    Ok. So the cost of voting [just in tripleo] would be
                    that a) [tripleo]<br>
                    infrastructure failures and b) breakage from other
                    projects - both<br>
                    things that can cause checks to fail, would stall
                    all tripleo landings<br>
                    until rectified, or until voting is turned off via a
                    change to config<br>
                    which makes this infra's problem.<br>
                    <br>
                    Hmm - so from a tripleo perspective, I think we're
                    ok with this -<br>
                    having a clear indication that 'this is ok' is
                    probably more important<br>
                    to us right now than the more opaque thing we have
                    now where we have<br>
                    to expand every jenkins comment to be sure.<br>
                    <br>
                    But- will infra be ok, if we end up having a
                    firedrill 'please make<br>
                    this nonvoting' change to propose?<br>
                    <div><br>
                      -Rob<br>
                      <br>
                      --<br>
                      Robert Collins <<a href="mailto:rbtcollins@hp.com" target="_blank">rbtcollins@hp.com</a>><br>
                      Distinguished Technologist<br>
                      HP Converged Cloud<br>
                      <br>
                      <br>
                      <br>
                      ------------------------------<br>
                      <br>
                    </div>
                    Message: 16<br>
                    Date: Sun, 16 Feb 2014 11:38:57 -0800<br>
                    From: Ravi Chunduru <<a href="mailto:ravivsn@gmail.com" target="_blank">ravivsn@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] [keystone] role of
                    Domain in VPC<br>
                            definition<br>
                    Message-ID:<br>
                           
                    <CAEgw6yuopjDfeF2vmAXtjiiA+Fz14=tbZcKV+m3eviLb=<a href="mailto:Xf5tQ@mail.gmail.com" target="_blank">Xf5tQ@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset="utf-8"<br>
                    <br>
                    I agree with JC that we need to pause and discuss
                    VPC model with in<br>
                    openstack before considering AWS compatibility. As
                    Subbu said, We need this<br>
                    discussion in Juno summit and get consensus.<br>
                    <br>
                    Thanks,<br>
                    -Ravi.<br>
                    <br>
                    <br>
                    On Sun, Feb 16, 2014 at 10:31 AM, Allamaraju, Subbu
                    <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>

                    wrote:<br>
                    <br>
                    > Harshad,<br>
                    ><br>
                    > This is great. At least there is consensus on
                    what it is and what it is<br>
                    > not. I would leave it to others to discuss
                    merits of a an AWS compat VPC<br>
                    > API for Icehouse.<br>
                    ><br>
                    > Perhaps this is a good topic to discuss at the
                    Juno design summit.<br>
                    ><br>
                    > Subbu<br>
                    ><br>
                    > On Feb 16, 2014, at 10:15 AM, Harshad Nakil
                    <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    > wrote:<br>
                    ><br>
                    > > As said I am not disagreeing with you or
                    Ravi or JC. I also agree that<br>
                    > > Openstack VPC implementation will benefit
                    from these proposals.<br>
                    > > What I am saying is it is not required AWS
                    VPC API compatibility at<br>
                    > > this point.  Which is what our blueprint
                    is all about. We are not<br>
                    > > defining THE "VPC".<br>
                    <div>><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      <br>
                      <br>
                      <br>
                    </div>
                    --<br>
                    Ravi<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2ef6cc51/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2ef6cc51/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 17<br>
                    Date: Sun, 16 Feb 2014 11:54:54 -0800<br>
                    From: Ravi Chunduru <<a href="mailto:ravivsn@gmail.com" target="_blank">ravivsn@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] VPC Proposal<br>
                    Message-ID:<br>
                            <<a href="mailto:CAEgw6ysbaY6-8w_VOme5mU1k29v0dy42mvuRkTsTR7XXKw6CMg@mail.gmail.com" target="_blank">CAEgw6ysbaY6-8w_VOme5mU1k29v0dy42mvuRkTsTR7XXKw6CMg@mail.gmail.com</a>><br>
                    Content-Type: text/plain; charset="utf-8"<br>
                    <br>
                    IMO, VPC means to have managed set of resources not
                    just limited to<br>
                    networks but also projects.<br>
                    I feel its not about incrementally starting with AWS
                    compatibility, But<br>
                    doing it right with AWS compatibility into
                    consideration.<br>
                    <br>
                    Thanks,<br>
                    -Ravi.<br>
                    <br>
                    <br>
                    On Sun, Feb 16, 2014 at 8:47 AM, Harshad Nakil<br>
                    <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>>wrote:<br>
                    <br>
                    > Comments Inline<br>
                    ><br>
                    > Regards<br>
                    > -Harshad<br>
                    ><br>
                    ><br>
                    > On Sat, Feb 15, 2014 at 11:39 PM, Allamaraju,
                    Subbu <<a href="mailto:subbu@subbu.org" target="_blank">subbu@subbu.org</a>>wrote:<br>
                    ><br>
                    >> Harshad,<br>
                    >><br>
                    >> Curious to know if there is a broad
                    interest in an AWS compatible API in<br>
                    >> the community?<br>
                    ><br>
                    ><br>
                    > We started looking at this as some our
                    customers/partners were interested<br>
                    > in get AWS API compatibility. We have this
                    blueprint and code review<br>
                    > pending for long time now. We will know based
                    on this thread wether the<br>
                    > community is interested. But I assumed that
                    community was interested as the<br>
                    > blueprint was approved and code review has no
                    -1(s) for long time now.<br>
                    ><br>
                    ><br>
                    >> To clarify, a clear incremental path from
                    an AWS compatible API to an<br>
                    >> OpenStack model is not clear.<br>
                    >><br>
                    ><br>
                    > In my mind AWS compatible API does not need new
                    openstack model. As more<br>
                    > discussion happen on JC's proposal and
                    implementation becomes clear we will<br>
                    > know how incremental is the path. But at high
                    level there two major<br>
                    > differences<br>
                    > 1. New first class object will be introduced
                    which effect all components<br>
                    > 2. more than one project can be supported
                    within VPC.<br>
                    > But it does not change AWS API(s). So even in
                    JC(s) model if you want AWS<br>
                    > API then we will have to keep VPC to project
                    mapping 1:1, since the API<br>
                    > will not take both VPC ID and project ID.<br>
                    ><br>
                    ><br>
                    <br>
                    <br>
                    <br>
                    > As more users want to migrate from AWS or IaaS
                    providers who want compete<br>
                    > with AWS should be interested in this
                    compatibility.<br>
                    ><br>
                    > There also seems to be terminology issue here
                    Whats is definition of "VPC"<br>
                    > if we assume what AWS implements is "VPC"<br>
                    > then what JC is proposing "VOS" or "VDC"
                    (virtual openstack or virtual DC)<br>
                    > as all or most of current openstack features
                    are available to user in  this<br>
                    > new Abstraction. I actually like this new
                    abstraction.<br>
                    ><br>
                    ><br>
                    >> Subbu<br>
                    >><br>
                    >> On Feb 15, 2014, at 10:04 PM, Harshad Nakil
                    <<a href="mailto:hnakil@contrailsystems.com" target="_blank">hnakil@contrailsystems.com</a>><br>
                    >> wrote:<br>
                    >><br>
                    >> ><br>
                    >> > I agree with problem as defined by you
                    and will require more<br>
                    >> fundamental changes.<br>
                    >> > Meanwhile many users will benefit from
                    AWS VPC api compatibility.<br>
                    <div>>><br>
                      >><br>
                      >>
                      _______________________________________________<br>
                      >> OpenStack-dev mailing list<br>
                      >> <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      >> <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      >><br>
                      ><br>
                      ><br>
                      >
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      ><br>
                      ><br>
                      <br>
                      <br>
                    </div>
                    --<br>
                    Ravi<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/745d6d7d/attachment-0001.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/745d6d7d/attachment-0001.html</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 18<br>
                    Date: Sun, 16 Feb 2014 12:08:15 -0800<br>
                    From: Vishvananda Ishaya <<a href="mailto:vishvananda@gmail.com" target="_blank">vishvananda@gmail.com</a>><br>
                    <div>To: "OpenStack Development Mailing
                      List (not for usage questions)"<br>
                              <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    Subject: Re: [openstack-dev] OpenStack-dev Digest,
                    Vol 22, Issue 39<br>
                    Message-ID: <<a href="mailto:91C14EC4-02F8-4DFC-9145-08BE2DA249AD@gmail.com" target="_blank">91C14EC4-02F8-4DFC-9145-08BE2DA249AD@gmail.com</a>><br>
                    Content-Type: text/plain; charset="windows-1252"<br>
                    <br>
                    <br>
                    On Feb 15, 2014, at 4:36 AM, Vinod Kumar Boppanna
                    <<a href="mailto:vinod.kumar.boppanna@cern.ch" target="_blank">vinod.kumar.boppanna@cern.ch</a>>

                    wrote:<br>
                    <br>
                    ><br>
                    > Dear Vish,<br>
                    ><br>
                    > I completely agree with you. Its like a trade
                    off between getting re-authenticated (when in a
                    hierarchy user has different roles at different
                    levels) or parsing the entire hierarchy till the
                    leaf and include all the roles the user has at each
                    level in the scope.<br>
                    ><br>
                    > I am ok with any one (both has some advantages
                    and dis-advantages).<br>
                    ><br>
                    > But one point i didn't understand why should we
                    parse the tree above the level where the user gets
                    authenticated (as you specified in the reply). Like
                    if user is authenticated at level 3, then do we mean
                    that the roles at level 2 and level 1 also should be
                    passed?<br>
                    > Why this is needed? I only see either we pass
                    only the role at the level the user is getting
                    authenticated or pass the roles at the level till
                    the leaf starting from the level the user is getting
                    authenticated.<br>
                    <br>
                    <br>
                    This is needed because in my proposed model roles
                    are inherited down the heirarchy. That means if you
                    authenticate against ProjA.ProjA2 and you have a
                    role like ?netadmin? in ProjA, you will also have it
                    in ProjA2. So it is necessary to walk up the tree to
                    find the full list of roles.<br>
                    <br>
                    Vish<br>
                    <br>
                    ><br>
                    > Regards,<br>
                    > Vinod Kumar Boppanna<br>
                    > ________________________________________<br>
                    > Message: 21<br>
                    > Date: Fri, 14 Feb 2014 10:13:59 -0800<br>
                    > From: Vishvananda Ishaya <<a href="mailto:vishvananda@gmail.com" target="_blank">vishvananda@gmail.com</a>><br>
                    <div>> To: "OpenStack Development
                      Mailing List (not for usage questions)"<br>
                      >        <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    </div>
                    <div>> Subject: Re: [openstack-dev]
                      Hierarchicical Multitenancy Discussion<br>
                    </div>
                    > Message-ID: <<a href="mailto:4508B18F-458B-4A3E-BA66-22F9FA47EAC0@gmail.com" target="_blank">4508B18F-458B-4A3E-BA66-22F9FA47EAC0@gmail.com</a>><br>
                    > Content-Type: text/plain;
                    charset="windows-1252"<br>
                    ><br>
                    > Hi Vinod!<br>
                    ><br>
                    > I think you can simplify the roles in the
                    hierarchical model by only passing the roles for the
                    authenticated project and above. All roles are then
                    inherited down. This means it isn?t necessary to
                    pass a scope along with each role. The scope is just
                    passed once with the token and the project-admin
                    role (for example) would be checking to see that the
                    user has the project-admin role and that the
                    project_id prefix matches.<br>
                    ><br>
                    > There is only one case that this doesn?t
                    handle, and that is when the user has one role (say
                    member) in ProjA and project-admin in ProjA2. If the
                    user is authenticated to ProjA, he can?t do
                    project-adminy stuff for ProjA2 without
                    reauthenticating. I think this is a reasonable
                    sacrifice considering how much easier it would be to
                    just pass the parent roles instead of going through
                    all of the children.<br>
                    ><br>
                    > Vish<br>
                    ><br>
                    <div>>
                      _______________________________________________<br>
                      > OpenStack-dev mailing list<br>
                      > <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      > <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      <br>
                    </div>
                    <div>-------------- next part
                      --------------<br>
                      A non-text attachment was scrubbed...<br>
                      Name: signature.asc<br>
                      Type: application/pgp-signature<br>
                    </div>
                    <div>Size: 455 bytes<br>
                      Desc: Message signed with OpenPGP using GPGMail<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/7c320704/attachment-0001.pgp" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/7c320704/attachment-0001.pgp</a>><br>


                    <br>
                    ------------------------------<br>
                    <br>
                    Message: 19<br>
                    Date: Sun, 16 Feb 2014 16:20:52 -0500<br>
                    From: Mike Spreitzer <<a href="mailto:mspreitz@us.ibm.com" target="_blank">mspreitz@us.ibm.com</a>><br>
                    To: "OpenStack Development Mailing List \(not for
                    usage questions\)"<br>
                            <<a href="mailto:openstack-dev@lists.openstack.org" target="_blank">openstack-dev@lists.openstack.org</a>><br>
                    Subject: Re: [openstack-dev] heat run_tests.sh fails
                    with one huge<br>
                            line    of      output<br>
                    Message-ID:<br>
                            <<a href="mailto:OF81356D12.13A4D038-ON85257C81.0073FA5A-85257C81.00754456@us.ibm.com" target="_blank">OF81356D12.13A4D038-ON85257C81.0073FA5A-85257C81.00754456@us.ibm.com</a>><br>
                    Content-Type: text/plain; charset="us-ascii"<br>
                    <br>
                    Kevin, I changed no code, it was a fresh DevStack
                    install.<br>
                    <br>
                    Robert Collins <<a href="mailto:robertc@robertcollins.net" target="_blank">robertc@robertcollins.net</a>>

                    wrote on 02/16/2014 05:33:59<br>
                    AM:<br>
                    > A) [fixed in testrepository trunk] the output
                    from subunit.run<br>
                    > discover .... --list is being shown verbatim
                    when an error happens,<br>
                    > rather than being machine processed and the
                    test listings elided.<br>
                    ><br>
                    > To use trunk - in your venv:<br>
                    > bzr branch lp:testrepository<br>
                    > pip install testrepository<br>
                    ><br>
                    > B) If you look at the end of that wall of text
                    you'll see 'Failed<br>
                    > imports' in there, and the names after that are
                    modules that failed<br>
                    > to import - for each of those if you try to
                    import it in python,<br>
                    > you'll find the cause, and there's likely just
                    one cause.<br>
                    <br>
                    Thanks Robert, I tried following your leads but got
                    nowhere, perhaps I<br>
                    need a few more clues.<br>
                    <br>
                    I am not familiar with bzr (nor baz), and it wasn't
                    obvious to me how to<br>
                    fit that into my workflow --- which was:<br>
                    (1) install DevStack<br>
                    (2) install libmysqlclient-dev<br>
                    (3) install flake8<br>
                    (4) cd /opt/stack/heat<br>
                    (5) ./run_tests.sh<br>
                    <br>
                    I guessed that your (A) would apply if I use a venv
                    and go between (1) the<br>
                    `python tools/install_venv.py` inside run_tests.sh
                    and (2) the invocation<br>
                    inside run_tests.sh of its run_tests function.  So I
                    manually invoked<br>
                    `python tools/install_venv.py`, then entered that
                    venv, then issued your<br>
                    commands of (A) (discovered I needed to install bzr
                    and did so), then<br>
                    exited that venv, then invoked heat's `run_tests -V
                    -u` to use the venv<br>
                    thus constructed.  It still produced one huge line
                    of output.  Here I<br>
                    attach a typescript of that:<br>
                    <br>
                    <br>
                    <br>
                    You will see that the huge line still ends with
                    something about import<br>
                    error, and now lists one additional package ---<br>
                    heat.tests.test_neutron_firewalld.  I then tried
                    your (B), testing manual<br>
                    imports.   All worked except for the last, which
                    failed because there is<br>
                    indeed no such thing (why is there a spurrious 'd'
                    at the end of the<br>
                    package name?).  Here is a typescript of that:<br>
                    <br>
                    <br>
                    <br>
                    Thanks,<br>
                    Mike<br>
                    <div>-------------- next part
                      --------------<br>
                      An HTML attachment was scrubbed...<br>
                    </div>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment.html" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment.html</a>><br>


                    -------------- next part --------------<br>
                    An embedded and charset-unspecified text was
                    scrubbed...<br>
                    Name: testlog.txt<br>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment.txt" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment.txt</a>><br>


                    -------------- next part --------------<br>
                    An embedded and charset-unspecified text was
                    scrubbed...<br>
                    Name: testlog2.txt<br>
                    URL: <<a href="http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment-0001.txt" target="_blank">http://lists.openstack.org/pipermail/openstack-dev/attachments/20140216/2f7188ae/attachment-0001.txt</a>><br>


                    <br>
                    ------------------------------<br>
                    <div><br>
                      _______________________________________________<br>
                      OpenStack-dev mailing list<br>
                      <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                      <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      <br>
                      <br>
                    </div>
                    End of OpenStack-dev Digest, Vol 22, Issue 45<br>
                    *********************************************<br>
                    <div>
                      <div><br>
                        _______________________________________________<br>
                        OpenStack-dev mailing list<br>
                        <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
                        <a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
                      </div>
                    </div>
                  </blockquote>
                </div>
                <br>
                <br clear="all">
                <div><br>
                </div>
                -- <br>
                <div dir="ltr">------------------------------------------<br>
                  Telles Mota Vidal Nobrega
                  <div>Bsc in Computer Science at UFCG<br>
                    Software Engineer at PulsarOpenStack Project -
                    HP/LSD-UFCG</div>
                </div>
              </div>
              <br>
              <fieldset></fieldset>
              <br>
              <pre>_______________________________________________
OpenStack-dev mailing list
<a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a>
</pre>
            </blockquote>
            <br>
          </div>
          _______________________________________________<br>
          OpenStack-dev mailing list<br>
          <a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
        </blockquote>
      </div>
      <br>
      <br>
      <fieldset></fieldset>
      <br>
      <pre>_______________________________________________
OpenStack-dev mailing list
<a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a>
</pre>
    </blockquote>
    <br>
  </div>

_______________________________________________<br>OpenStack-dev mailing list<br><a href="mailto:OpenStack-dev@lists.openstack.org" target="_blank">OpenStack-dev@lists.openstack.org</a><br><a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>

</blockquote></div><br></div><br>_______________________________________________<br>
OpenStack-dev mailing list<br>
<a href="mailto:OpenStack-dev@lists.openstack.org">OpenStack-dev@lists.openstack.org</a><br>
<a href="http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev" target="_blank">http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br><div dir="ltr">------------------------------------------<br>Telles Mota Vidal Nobrega<div>Bsc in Computer Science at UFCG<br>Software Engineer at PulsarOpenStack Project - HP/LSD-UFCG</div>

</div>
</div>