[Openstack-operators] Kilo virtual machine stop to aquire dhcp address :-(

Ignazio Cassano ignaziocassano at gmail.com
Thu Jul 23 17:35:50 UTC 2015


Many thanks
Il giorno 23/lug/2015 19:26, "Salvatore Orlando" <salv.orlando at gmail.com>
ha scritto:

> More comments inline.
>
> Salvatore
>
> On 23 July 2015 at 16:52, Ignazio Cassano <ignaziocassano at gmail.com>
> wrote:
>
>> Hi Salvatore, thank you for your answer
>> When dhcp does not release an ip address the virtual machines cannot ping
>> the gateway also if you assign them
>> a static address.
>>
>
>> I deleted  all subnets and I created only one subnet : now the
>> virtualmachines installed from the standard cirros image do not aquire ip
>> address but if you assign them the ip address displayed on the dashboard
>> they can ping the gateway.
>> Then, if I reboot one virtualmachine , it aquires the ip address,
>> hostname and the mtu from dhcp. It is a very strange behaviour.
>>
>
> This might indicate delays in setting up configurations in the network
> node.
> Rather than rebooting the VM you can just try to send more DHCPREQUEST
> with udhcpc from the cirros instance.
> If it eventually receives an address, a delay on the network node is the
> most likely issue.
>
> Nevertheless, this would hardly explain why you are unable to ping the
> gateway even with static addresses, as the router interface should be
> already up and configured (you should indeed be able to see it in the
> qrouter-<router_id> namespace and also see a veth pair connecting that
> interface to the ovs bridge.
>
>
>>
>> A question (I do not know openvswitch):
>> I saw openvswitch creates several ports either on network node or
>> computer node.
>> Some of these ports have a tag number (vlan ?).
>> Must the port assigned to the gateway and the dhcp server for a subnet
>> have the same tag number of the port assigned to a virtualmachine in the
>> same subnet ?
>>
>
> The tag number is exclusively local to the host. Its purpose is to isolate
> traffic for different tenants within the same host. Traffic for the same
> logical network might be given distinct tags on distinct hosts.
> I hope I got your question right!
>
>
>> Regards
>>
>> 2015-07-23 16:58 GMT+02:00 Salvatore Orlando <salv.orlando at gmail.com>:
>>
>>> When you manually setup IP addressing on the VMs, are they able to ping
>>> each other?
>>> The issue you are describing is compatible with a scenario where both
>>> the dhcp agent and the l3 agent on the same "network node", and that node
>>> gets disconnected from the rest of the fabric.
>>>
>>> This could be for several reasons. At the control plane level, the l2
>>> agent might be not operating correctly (even if its reports an "up"
>>> status); at the data plane level instead l3 connectivity with the network
>>> node might be interrupted. The l2 agent log on the network node might be
>>> able to answer the question concerning whether dhcp and router interface
>>> are being wired correctly.
>>>
>>> Salvatore
>>>
>>>
>>> On 23 July 2015 at 09:49, Ignazio Cassano <ignaziocassano at gmail.com>
>>> wrote:
>>>
>>>> Hi guys,
>>>> I installed openstack kilo last week and it worked fine for some days.
>>>> On yesterday my virtual machines stop to aquire dhcp address and if I
>>>> set address
>>>> manually they cannot ping either dhcp server or gateway.
>>>>
>>>> My environment is made up of:
>>>>
>>>> 1 controller node
>>>> 1 neutron netork node
>>>> 1 kvm  computing node
>>>> 2 object storage nodes
>>>> 2 lvm iscsi block storage nodes
>>>>
>>>> If I run "neutron agent-list" on controller node it reports the
>>>> following lines:
>>>>
>>>> +--------------------------------------+--------------------+-----------------+-------+----------------+---------------------------+
>>>> | id                                   | agent_type         |
>>>> host            | alive | admin_state_up | binary                    |
>>>>
>>>> +--------------------------------------+--------------------+-----------------+-------+----------------+---------------------------+
>>>> | 7e18abbf-8c6b-4463-9833-44b45dc7fec8 | Open vSwitch agent |
>>>> OpStk-NetNode   | :-)   | True           | neutron-openvswitch-agent |
>>>> | 7e67678a-5eb1-4d8b-9a44-f084187ddb03 | Open vSwitch agent |
>>>> OpStk-CompNode1 | :-)   | True           | neutron-openvswitch-agent |
>>>> | 87ebbf56-8405-4b0d-84b4-a3ad7e3c7a0c | L3 agent           |
>>>> OpStk-NetNode   | :-)   | True           | neutron-l3-agent          |
>>>> | ac6d21c3-ab38-451e-9393-9d38ea1058ec | DHCP agent         |
>>>> OpStk-NetNode   | :-)   | True           | neutron-dhcp-agent        |
>>>> | ebb2b82e-1cba-4a69-b943-bf69e7c8a478 | Metadata agent     |
>>>> OpStk-NetNode   | :-)   | True           | neutron-metadata-agent    |
>>>>
>>>> +--------------------------------------+--------------------+-----------------+-------+----------------+---------------------------+
>>>>
>>>> Could anyone help me ?
>>>>
>>>> Many thanks and regards
>>>> Ignazio
>>>>
>>>>
>>>> _______________________________________________
>>>> OpenStack-operators mailing list
>>>> OpenStack-operators at lists.openstack.org
>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>>>>
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20150723/081ab3a4/attachment.html>


More information about the OpenStack-operators mailing list