Open Stack

hi Vikas Choudhary, when ovs-agent service recover(ovs-agent process restart), the dhcp port will not re-binding successfully?





At 2015-08-22 14:26:08, "Vikas Choudhary" <choudharyvikas16 at gmail.com> wrote:

Hi Everybody,


I want to discuss on https://bugs.launchpad.net/neutron/+bug/1348589.This is there for more than a year and no discussion i could find on this.


Scenario:
ovs-agent is down and then a network and subnet under this newly created network are created using cli. No error visible to user, but following irregularities are found.


Discrepancies:
1. neutron port-show <dhcp-port> shows:
     binding:vif_type      | binding_failed
2. Running "ovs-ofctl dump-flows br-tun", no of-flow got added
3. Running "ovs-vsctl show br-int", no tag for dhcp-port. 


neutron db will have all the attributes required to retry vif binding.My query is when should we trigger this rebinding.Two approaches i could think of are:
1> At neutron server restart, for all ports with vif_type as  "binding_failed" plugins/ml2/drivers/mech_agent.bind_port can be invoked as a sync up activity.


2> In neutron port update api, http://developer.openstack.org/api-ref-networking-v2-ext.html , could be enhanced to receive vif binding related options also and then eventually plugins/ml2/drivers/mech_agent.bind_port can be invoked.Corresponding changes will be made to 'port update cli' also.


Please suggest.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150822/f45d4807/attachment.html>