[Openstack-operators] Instances not getting DHCP Lease

Curtis serverascode at gmail.com
Sat Sep 2 21:19:21 UTC 2017

On Sat, Sep 2, 2017 at 1:57 AM, Divneet Singh <divneetsingh.ml at gmail.com> wrote:
>>What I usually do, and this might not be easiest/best, is, given you
>>do have console access, to manually set the IP in the cirrus node.
>>Then I enter the ip name space for the dhcp server on the network node
>>(sudo ip netns exec qdhcp-<thenetwork> bash), and see if I can ping
>>the cirrus node and vice-versa, things like that. Making sure to add a
>>security group rule to at least allow ICMP. See if you can at least
>>get an arp entry for the virtual machine's IP.
>>Usually if there is no connectivity at all it's b/c a vlan is missing
>>for the physical interface on the switch (not sure how your deployment
>>is setup), or the also neutron configuration settings are incorrect in
>>terms of bridge mappings and such, usually something simple like that.
>>Another option is to reboot or retry dhcp on the cirrus vm and tcpdump
>>on the physical interface the dhcp server is on and watch for dhcp
>>traffic, and you could do that on the compute node as well, looking
>>for both request and response, again to validate basic connectivity.
>>If you still are having issues, let us know how you deployed openstack
> (distro?) and what kind of network you are using (vxlan, vlan?),
> whether it's linux bridge or ovs, things like that.
> I have deployed the nodes using a virtual box & set up nat 2 nat networks ,
> configured the neutron service according to self service network guide . I
> have doubled checked the neutron configuration . I do agree to fact the

A bit might have gotten cut off there. :)

At any rate, I'm not sure what nat 2 nat networks are in virtual box,
but I'm going to guess that you have a problem in your virtual box
networking setup. You'll have to investigate those networks and make
sure they are setup in a way that will work with your openstack
deployment. If you are just trying out openstack, perhaps a single
node devstack install would be a good place to start and then move to
more advanced deployments from there?


> Although i can't make out whether there is connectivity between the nodes.
> I get the following output when i run the tcp dump on the controller node .
> listening on enp0s3, link-type EN10MB (Ethernet), capture size 262144 bytes
> 12:48:26.488533 IP > Flags [P.], seq
> 1408700411:1408700519, ack 23895, win 40080, length 108
> 12:48:26.488719 IP > Flags [P.], seq
> 108:144, ack 1, win 40080, length 36
> Output on the compute node .
> listening on enp0s3, link-type EN10MB (Ethernet), capture size 262144 bytes
> 12:59:21.558954 IP > Flags [P.], seq
> 1404672109:1404672297, ack 639164055, win 312, options [nop,nop,TS val
> 677134 ecr 715780], length 188
> 12:59:21.559252 IP > Flags [.], ack
> 188, win 1444, options [nop,nop,TS val 715791 ecr 677134], length 0

Blog: serverascode.com

More information about the OpenStack-operators mailing list