Instance does'n get IP.

asked 2018-05-28 03:07:59 -0500

sima_attar2003

3 ●1 ●3 ●4

Hello, I know it may seem duplicate question, but in fact its not. I have a 4 node openstack in a real deployment, one controller and 3 computes. I have added another 6 compute nodes. On new compute nodes, instance launches successfully, and in dashboard it shows that instance gets IP, but inside the instance there is no IP assigned to it, "ip a" shows no IP and when I ping it through router, it in not reachable. inside the instance obviously "Network is unreachable" too, since no IP is assigned to it. The weird thing is that while launching the instance, there is no error log in controller and compute. The way I used for adding the new computes, is that I installed packages: openstsack-nova-compute, openstack-selinux, openstack-neutron, openstack-neutron-openvswitch on new computes. then I copy configuration files : nova.conf , neutron.conf, openvswitch-agent.ini from a properly working compute node, to the new ones. then I changes IP addresses and hostnames, plus chown of files, accordingly. after that start and enable the services.

Any help would be extremely appreciated.

edit retag flag offensive close merge delete

add a comment

0

answered 2018-05-28 06:51:08 -0500

Bernd Bausch

6319 ●40 ●74 ●146 https://www.linkedin.c...

updated 2018-05-28 06:52:09 -0500

When an instance gets no IP address, this could be due to a problem with DHCP, or the instance, or a problem with the instance's connection to the network, or DHCP connection to the network.

Since you seem to have access to the instances via their console, try setting their IP addresses and routing table manually. If this gives you connectivity, the problem is on the DHCP side.

Check the obvious: Does the instance issue DHCP requests? Is DHCP enabled on the networks to which the instances are connected?

How did you configure the tenant networks? Perhaps you followed the installation tutorial. In any case, double-check the content of the configuration files. A simple typo, e.g. "vlan" instead of "vxlan", can break everything without necessarily leaving a trace in the log file.

You could also post the relevant Neutron config files (neutron.conf, ML2 config, DHCP config, Neutron agent config depending on which mechanism driver you use) and details of your tenant networks and subnets. http://paste.openstack.org is a good place for that.