Ask Your Question
1

Why would DHCP intermittently assign 2 IP addresses?

asked 2016-06-16 03:19:06 -0500

mickt gravatar image

I have a mitaka installation running in docker containers; dockerisation should be irrelevant. I notice that the DHCP server intermittently assigns 2 IP addresses to instances. Anyone know why this might be? I also currently have HA enabled and this too should be irrelevant as I see the same behaviour with HA disabled.

I boot instances as follows:

nova boot --flavor m1.tiny --image cirros --nic net-id=f6a73223-5dbe-4bc6-94e6-aaca898a844d --security-group default --key-name cloud test

nova list:

+--------------------------------------+-------+--------+------------+-------------+------------------------------------+

| ID | Name | Status | Task State | Power State | Networks |

+--------------------------------------+-------+--------+------------+-------------+------------------------------------+

| b4aee59e-c08b-49d2-8ae9-33e9b6708a49 | test | ACTIVE | - | Running | int-net=195.7.41.3, 10.196.229.201 |

| 4c697b2f-41d3-4456-b994-4d311f2d7785 | test1 | ACTIVE | - | Running | int-net=195.7.41.4, 195.7.41.5 |

| 75f34b58-3b38-4c19-b922-fff1a0e84002 | test2 | BUILD | spawning | NOSTATE | int-net=195.7.41.6 |

+--------------------------------------+-------+--------+------------+-------------+------------------------------------+

.

You can see that instance test1 has two IP addresses assigned from 195.7.41.0.

edit retag flag offensive close merge delete

Comments

1

On further test I notice that scheduling shows 1 IP & spawning shows 2:

e237380e-6812-40c7-98f0-06b629b9d709 | test1 | BUILD | scheduling | NOSTATE | int-net=195.7.41.5

e237380e-6812-40c7-98f0-06b629b9d709 | test1 | BUILD | spawning | NOSTATE | int-net=195.7.41.5, 195.7.41.6

mickt gravatar imagemickt ( 2016-06-16 09:19:26 -0500 )edit

We hit the same behavior after 2nd bridge was added to following config: /etc/neutron/plugins/ml2/openvswitch_agent.ini - bridge_mappings. when you specify --availability-zone nova:compute1 (AZ with only 1 node) the problem is gone - instance has only 1 expected IP.

jan-kaufman-w gravatar imagejan-kaufman-w ( 2016-06-20 09:35:08 -0500 )edit

We are also Facing same problem using Mitaka on Ubuntu with ODL plugin.

bsrgsit gravatar imagebsrgsit ( 2016-06-22 23:54:04 -0500 )edit

Alas, AZ do not resolve my issue.

mickt gravatar imagemickt ( 2016-07-14 09:28:54 -0500 )edit

We have the same issue on Mitaka with opensuse_Leap_42.1 and don't know how to fix it yet.

eblock gravatar imageeblock ( 2016-07-28 05:38:41 -0500 )edit

4 answers

Sort by ยป oldest newest most voted
1

answered 2016-06-20 09:27:32 -0500

Arteal gravatar image

Same happens to us, from time when we added second bridge mapping to :

/etc/neutron/plugins/ml2/openvswitch_agent.ini

And it happens only when you do not specify availability zone to specific node.

edit flag offensive delete link more

Comments

Sending availability zone solved problem for me too.

shital gravatar imageshital ( 2017-03-16 08:33:15 -0500 )edit
0

answered 2016-06-20 08:57:37 -0500

Hi Mickt,

we are experiencing the same thing on Ubuntu 14.04 w/ Cloud archive Mitaka. Have you solved the issue already?

Regs David

edit flag offensive delete link more

Comments

Not yet, it is low priority for me at moment but is on my to-do list.

mickt gravatar imagemickt ( 2016-07-29 02:29:48 -0500 )edit
0

answered 2017-06-02 04:00:00 -0500

zhangoic gravatar image

updated 2017-06-02 04:06:52 -0500

I hit this problem either. I don' think it has any relation with bridge-mapping. In my experice, I found that vm spawn has two scheduling process. In the first one, the tap interface in ovs has been created by nova-compute, and the port status in neutron db was active. But then, the compute node met qemu error (CPU feature were not found). The exception made nova-scheduler to reschedule. The second process would also create port and ovs interface by a new compute node. As a result, it has two port in neutron. After I closed the fault one, the problem was dissappeared. All new created vm had only one port.

edit flag offensive delete link more
0

answered 2017-09-05 05:02:47 -0500

Hello zhangoic,

Can you help me to fix same issue? Or can you tell me what changes was made to solve it?

Thanks.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

3 followers

Stats

Asked: 2016-06-16 03:19:06 -0500

Seen: 1,327 times

Last updated: Jun 02 '17