[Openstack-operators] Weird behaviour of dhcp agent

Alvise Dorigo alvise.dorigo at pd.infn.it
Thu Mar 27 12:42:46 UTC 2014

I’ve two network nodes running two dhcp agent instances. 
I’ve a tenant network dynamically binded to both dhcp agent.

If I create a new VM (cirros 0.3.1), the dhcp redundancy works as expected; if one dhcp agent is down, the other one correctly respond to the dhcp request of the new instance. 
Once the instance is UP I try to reboot (not terminate/re-create, just hard-reboot) the instance. It gets correctly the IP.

Then I move the dhcp agent from one network node to the other. If I reboot the VM it doesn’t get anymore the IP. I’ve to destroy it, re-create it and it will get correctly the IP.

Then, once an instances has got an IP from a certain DHCP agent, that instance seems to be forever bound to that dhcp agent… and it will use alway that dhcp across reboots.

To use a different dhcp agent, the instance must be re-created (after destroy).

Is that normal, or should I configure something ?


More information about the OpenStack-operators mailing list