<html>
  <head>
    <meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    Hi again,<br>
    <br>
    So the problem is now solved.<br>
    I hereby post the solution for people from the future.<br>
    <br>
    1. the ping between the compute and the controller was using an IP
    route. So the ping wasn't using only layer 2. This means that there
    was no DHCP request arriving to the network controller.<br>
    2. the hosts and the VMs should be in the same subnet<br>
    3. we needed to killall dnsmasq and restart nova-network<br>
    <br>
    tcpdump on br100 is useful to track dhcp requests. ARP tables are
    useful as well in order to be sure each host sees the other on layer
    2.<br>
    <br>
    thank you all,<br>
    <br>
    yours,<br>
    <br>
    michaël<br>
    <br>
    <br>
    <br>
    <pre class="moz-signature" cols="72">Michaël Van de Borne
R&D Engineer, SOA team, CETIC
Phone: +32 (0)71 49 07 45 Mobile: +32 (0)472 69 57 16, Skype: mikemowgli
<a class="moz-txt-link-abbreviated" href="http://www.cetic.be">www.cetic.be</a>, rue des Frères Wright, 29/3, B-6041 Charleroi
</pre>
    <br>
    Le 10/05/2012 15:31, Yong Sheng Gong a écrit :
    <blockquote
cite="mid:OF10022C0D.2F4FDD1F-ON482579FA.004A547F-482579FA.004A5483@cn.ibm.com"
      type="cite"><font face="Default Sans
        Serif,Verdana,Arial,Helvetica,sans-serif" size="2">
        <div>HI,<br>
          First you have to make sure the network between your control
          node's br100 and your compute node's br100 are connected. <br>
          and then can you show the output on control node:<br>
          ps -ef | grep dnsmasq<br>
          <div>brctl show<br>
            ifconfig<br>
            2. can you login to your vm by vnc to see the eth0
            configuration and then try to run udhcpc?<br>
            <br>
            Thanks<br>
          </div>
          <font color="#990099">-----openstack-bounces+gongysh=cn.ibm.com@lists.launchpad.net
            wrote: -----<br>
            <br>
          </font>
          <blockquote style="padding-right: 0px; padding-left: 5px;
            margin-left: 5px; border-left: 2px solid #000000;
            margin-right: 0px;">To: <a moz-do-not-send="true"
              class="moz-txt-link-rfc2396E"
              href="mailto:openstack@lists.launchpad.net">"openstack@lists.launchpad.net"</a>
            <a moz-do-not-send="true" class="moz-txt-link-rfc2396E"
              href="mailto:openstack@lists.launchpad.net"><openstack@lists.launchpad.net></a><br>
            From: Michaël Van de Borne <a moz-do-not-send="true"
              class="moz-txt-link-rfc2396E"
              href="mailto:michael.vandeborne@cetic.be"><michael.vandeborne@cetic.be></a><br>
            Sent by: <a moz-do-not-send="true"
              class="moz-txt-link-abbreviated"
              href="mailto:openstack-bounces+gongysh=cn.ibm.com@lists.launchpad.net">openstack-bounces+gongysh=cn.ibm.com@lists.launchpad.net</a><br>
            Date: 05/10/2012 09:03PM<br>
            Subject: [Openstack] Accessing VMs in Flat DHCP mode with
            multiple host<br>
            <br>
            <!--Notes ACF <meta http-equiv="content-type" content="text/html; ">-->
            Hello,<br>
            <br>
            I'm running into troubles accessing my instances.<br>
            I have 3 nodes:<br>
            1. proxmox that virtualizes in KVM my controller node<br>
                1.1 the controller node (10.10.200.50) runs keystone,
            nova-api, network, scheduler, vncproxy and volumes but NOT
            compute as it is already a VM<br>
            2. glance in a physical node<br>
            3. compute in a physical node<br>
            <br>
            my nova.conf network config is:<br>
            <tt>--dhcpbridge_flagfile=/etc/nova/nova.conf<br>
              --dhcpbridge=/usr/bin/nova-dhcpbridge<br>
              --routing_source_ip=10.10.200.50<br>
              --libvirt_use_virtio_for_bridges=true<br>
              --network_manager=nova.network.manager.FlatDHCPManager<br>
              --public_interface=eth0<br>
              --flat_interface=eth1<br>
              --flat_network_bridge=br100<br>
              --fixed_range=192.168.200.0/24<br>
              --floating_range=10.10.200.0/24 <br>
              --network_size=256<br>
              --flat_network_dhcp_start=192.168.200.5<br>
              --flat_injected=False<br>
              --force_dhcp_release<br>
              --network_host=10.10.200.50</tt><br>
            <br>
            I even explicitly allows icmp and tcp port 22 traffic like
            this:<br>
            <font face="Courier New,Courier,monospace" size="3"><a
                moz-do-not-send="true"
                id="enabling-access-to-vms-on-the-compute-node">euca-authorize
                -P icmp -t -1:-1 default<br>
                euca-authorize -P tcp -p 22 default<br>
                <br>
              </a></font> <a moz-do-not-send="true"
              id="enabling-access-to-vms-on-the-compute-node">before
              setting these rules, I was getting 'Operation not
              permitted' when pinging the VM from the compute node.
              After setting these, I just get no output</a> at all (not
            even 'Destination Host Unreachable')<br>
            <a moz-do-not-send="true"
              id="enabling-access-to-vms-on-the-compute-node"><br>
              <br>
              The network was created like this:<br>
              nova-manage network create private
              --fixed_range_v4=192.168.200.0/24 --bridge=br100
              --bridge_interface=eth1 --num_networks=1
              --network_size=256<br>
              <br>
              However I cannot ping or ssh my instances once they're
              active. I have already set up such an Essex environment
              but the controller node was physical. Morevover, every
              examples in the doc presents a controller node that runs
              nova-compute.<br>
              <br>
              So I'm wondering if either:<br>
              - having the controller in a VM<br>
              - or not running compute on the controller<br>
              would prevent things to work properly.<br>
              <br>
              What can I check? iptables? is dnsmasq unable to give the
              VM an address? <br>
              <br>
              I'm running out of ideas. Any suggestion would be highly
              appreciated.<br>
              <br>
              Thank you,<br>
              <br>
              michaël<br>
            </a><br>
            <br>
            <br>
            <br>
            <font face="Courier New,Courier,monospace" size="3">-- <br>
              Michaël Van de Borne<br>
              R&D Engineer, SOA team, CETIC<br>
              Phone: +32 (0)71 49 07 45 Mobile: +32 (0)472 69 57 16,
              Skype: mikemowgli<br>
              <a moz-do-not-send="true" class="moz-txt-link-abbreviated"
                href="http://www.cetic.be">www.cetic.be</a>, rue des
              Frères Wright, 29/3, B-6041 Charleroi<br>
            </font> <font face="Courier New,Courier,monospace" size="3">_______________________________________________<br>
              Mailing list: <a moz-do-not-send="true"
                href="https://launchpad.net/%7Eopenstack">https://launchpad.net/~openstack</a><br>
              Post to     : <a moz-do-not-send="true"
                class="moz-txt-link-abbreviated"
                href="mailto:openstack@lists.launchpad.net">openstack@lists.launchpad.net</a><br>
              Unsubscribe : <a moz-do-not-send="true"
                href="https://launchpad.net/%7Eopenstack">https://launchpad.net/~openstack</a><br>
              More help   : <a moz-do-not-send="true"
                href="https://help.launchpad.net/ListHelp">https://help.launchpad.net/ListHelp</a><br>
            </font>
          </blockquote>
          <br>
        </div>
      </font>
      <br>
    </blockquote>
  </body>
</html>