Feature #5102

Open vSwitch VMs lost connection after hyp network restart

Added by Alvaro Simon almost 3 years ago.

Status:PendingStart date:04/11/2017
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:Drivers - Network
Target version:-
Resolution: Pull request:

Description

Dear OpenNebula developers

This issue is related with https://forum.opennebula.org/t/centos-7-network-restart-and-open-vswitch/2999

The problem is when the network service is restarted by any reason in the hypervisor, OVS also removes and regenerates the virtual bridge and all the connected ports. The VMs lost the network connection at this point.
The workaround at this moment is to change the VM status or migrate the VM to a different host, this way the ovsctl command is executed again by oneadmin user and the network and VLANs are again assigned to the VM as usual.

I do not know if there is a way to "force" to execute the virtual network prolog again to regenerate the missing network ports after a service network restart in the hyps. There are several options like a) monitoring from ONE server if the VM ports are not there or/and b) provide a command to regenerate the missing ports running the prolog manually. This OVS "feature" also affects more cloud frameworks like openstack, in that case the ovs network is managed by a daemon running in the hyp which checks the available OVS ports all the time. But probably this is not the best option and also breaks the OpenNebula KISS principle.

Any suggestion regarding this?

Cheers and thanks!
Álvaro

Also available in: Atom PDF