Revision history [back]

click to hide/show revision 1
initial version

I met the same problem. And it seems the code tried to seek the record with agent_type=l3 and host=XX (XX was the node where the instance located) in database. However, the installation guide only installed the l3_agent in controller node.

root@controller:~/openstack/pike/controller/script# openstack network agent list --agent-type l3 +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+ | ID | Agent Type | Host | Availability Zone | Alive | State | Binary | +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+ | db00614d-5172-4057-b688-87b137fe3033 | L3 agent | controller | nova | :-) | UP | neutron-l3-agent | +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+

Then I just installed neutron-l3-agent in all of the compute nodes.

root@cactl:~/openstack/pike/controller/script# openstack network agent list --agent-type l3 +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+ | ID | Agent Type | Host | Availability Zone | Alive | State | Binary | +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+ | 64471b43-fc9f-4991-9fe4-d314551f803a | L3 agent | cac01 | nova | :-) | UP | neutron-l3-agent | | 67eec814-418b-41f2-80b3-cf134f5c1983 | L3 agent | cac02 | nova | :-) | UP | neutron-l3-agent | | 5b537e1b-d35d-4370-93a7-5a2b529acd02 | L3 agent | cac03 | nova | :-) | UP | neutron-l3-agent | | db00614d-5172-4057-b688-87b137fe3033 | L3 agent | controller | nova | :-) | UP | neutron-l3-agent | +--------------------------------------+------------+-------+-------------------+-------+-------+------------------+

And the problem seems to be disappeared.

It should be a bug in the code and the host seems to be "controller" instead of the name of one of the compute node.

Anyway, using the method I mentioned above seems to be a walk-around. But I'm not sure about the side affects.