Havana instance doesn't get IP [closed]

asked 2014-02-22 03:49:18 -0600

anonymous user


updated 2014-04-25 11:27:06 -0600

larsks gravatar image

Hello, I have a custom Havana configuraton based on CentOS 6.5, with epel and redo repos enabled (but with rabbit for message queues instead of qpid)

All seems good apart instance (that runs on one compute node) that doesn't get ip address from dhcp agent (that runs on one controller node).

I'm using neutron with OVS and vlan.

In messages of controller node where dnsmasq process has started I see sometimes nothing and sometimes only dhcpdiscover and dhcpoffer lines for the mac of the instance, but nothing more and at instance side no ip has been got.

Tried with Cirros, CentOS 6.4, Fedora 20 and precise cloud images but all manifest this problem.

At the end of the message you can find a link to a file with al the commands run to debug problem (I'm not sure if directly pasting here gets correct output). Eventually I can replicate the environment with packstack and 1 compute + 1 network node, if the resulting configuration is similar, so I can compare files and behaviours.

My configuration is like this

kernel is the latest CentOS: 2.6.32-431.el6.x86_64 (it seems it is newer than the latest rdo provided one that is kernel-2.6.32-358.123.2.openstack.el6.x86_64.rpm and so yum update comand picked it)

for example openstack and openvswitch packages from rdo such as openvswitch-1.11.0_8ce28d-1.el6ost.x86_64 openstack-neutron-2013.2.1-1.el6.noarch

controller node with Neutron Server, Neutron OVS plugin, Neutron DHCP Agent

From Linux services started point of view:

  • neutron-dhcp-agent
  • neutron-openvswitch-agent
  • neutron-ovs-cleanup
  • neutron-server
  • openstack-cinder-api
  • openstack-cinder-scheduler
  • openstack-glance-api
  • openstack-glance-registry
  • openstack-heat-api
  • openstack-heat-api-cfn
  • openstack-heat-engine
  • openstack-keystone
  • openstack-nova-api
  • openstack-nova-conductor
  • openstack-nova-consoleauth
  • openstack-nova-novncproxy
  • openstack-nova-scheduler
  • openstack-nova-xvpvncproxy
  • openvswitch

Compute with Nova Compute, Neutron OVS Agent

From Linux services point of view:

  • neutron-openvswitch-agent
  • neutron-ovs-cleanup
  • openstack-cinder-volume
  • openstack-nova-compute
  • openvswitch

mac of instance: fa:16:3e:4a:eb:2b

on controller node I see dnsmasq active and generated file such as:

[root@vcontrol03 ~]# cat /var/lib/neutron/dhcp/d5bca3a8-2255-4a13-b0a7-f0d644bc4670/host

but actually no message regarding dhcp in messages The servers actually are virtual machines inside a vSPhere 5.1 infra and on their eth1 adapter (used for br-ex) I have configured trunk port and promiscuous mode

Below link with output of many commands (openvswitch, neutron, iptables...) on controller and compute. because directly inside is scrambling...


edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by rbowen
close date 2014-06-20 14:42:23.412131


Are you still experiencing this problem? Let us know. Thanks!

larsks gravatar imagelarsks ( 2014-04-25 11:28:33 -0600 )edit

I have the same problem, use RDO packstack deploy havana based on CentOS 6.5, I'm using neutron with OVS and GRE. Everything looks fine, but instance doesn't get IP. Hope can get some help here.

aamw gravatar imageaamw ( 2014-04-25 16:43:16 -0600 )edit

@aamw, you probably want to open a new question. This will make it easier to ask you for details about your environment.

larsks gravatar imagelarsks ( 2014-04-25 20:03:38 -0600 )edit

Could you reproduce similar report Capture DHCP Requests and ACKs on tap-interface on Controller.
I have noticed that on working Two Node Havana Neutron OVS&GRE Cluster my ACK's udp checksums were reported bad by tcpdump -i tap-xxxxxxx -vv.
. However /var/log/neutron/dnsmasq.log doesn't report any errors working with VMs

dbaxps gravatar imagedbaxps ( 2014-04-27 02:07:26 -0600 )edit