Open vSwitch 3.8
This guide describes how to use the Open vSwitch network drives. They provide two indepent functionalities that can be used together: network isolation using VLANs, and network filtering using OpenFlow. Each Virtual Network interface will receive a VLAN tag enabling network isolation. Other traffic attributes that may be configured through Open vSwitch are not modified.
The VLAN id will be the same for every interface in a given network, calculated by adding a constant to the network id. It may also be forced by specifying an VLAN_ID parameter in the Virtual Network template.
The network filtering functionality is very similar to the Firewall drivers, with a few limitations discussed below.
This driver requires Open vSwitch to be installed on each OpenNebula Host. Follow the resources specified in hosts_configuration to install it.
Integrating OpenNebula with Open vSwitch brings a long list of benefits to OpenNebula, read Open vSwitch Features to get a hold on these features.
This guide will address the usage of VLAN tagging and OpenFlow filtering of OpenNebula Virtual Machines. On top of that any other Open vSwitch feature may be used, but that's outside of the scope of this guide.
sudoers
file must be configured so oneadmin
can execute ovs_vsctl
in the hosts.To enable this driver, use ovswitch as the Virtual Network Manager driver parameter when the hosts are created with the onehost command:
<xterm> $ onehost create host01 im_kvm vmm_kvm tm_shared ovswitch </xterm>
Action | Description |
---|---|
Pre | - |
Post | Performs the appropriate Open vSwitch commands to tag the virtual tap interface. |
Clean | It doesn't do anything. The virtual tap interfaces will be automatically discarded when the VM is shut down. |
The driver will be automatically applied to every Virtual Machine deployed in the Host. Only the virtual networks with the attribute VLAN
set to YES
will be isolated. There are no other special attributes required.
NAME = "ovswitch_net" TYPE = "fixed" BRIDGE = vbr1 VLAN = "YES" VLAN_ID = 50 # optional LEASES = ...
VLAN_ID
parameter in the network template. In that scenario, any user may be able to connect to another network with the same network id. Techniques to avoid this are explained under the Tuning & Extending section.
The first rule that is always applied when using the Open vSwitch drivers is the MAC-spoofing rule, that prevents any traffic coming out of the VM if the user changes the MAC address.
The firewall directives must be placed in the network section of the Virtual Machine template. These are the possible attributes:
iptables_range
: Doesn't permit access to the VM through the specified ports in the TCP protocol. Superseded by WHITE_PORTS_TCP if defined.iptables_range
: Doesn't permit access to the VM through the specified ports in the UDP protocol. Superseded by WHITE_PORTS_UDP if defined.
iptables_range
: a list of ports separated by commas, e.g.: 80,8080
. Currently no ranges are supporteg, e.g.: 5900:6000
is not supported.
Example:
NIC = [ NETWORK_ID = 3, BLACK_PORTS_TCP = "80, 22", ICMP = drop ]
/var/lib/one/remotes
directory won't be effective in the Hosts until you execute, as oneadmin:
<xterm> oneadmin@frontend $ onehost sync </xterm>
This way in the next monitoring cycle the updated files will be copied again to the Hosts.
The vlan id is calculated by adding the network id to a constant defined in /var/lib/one/remotes/vnm/OpenNebulaNetwork.rb
. You can customize that value to your own needs:
CONF = { :start_vlan => 2 }
You can either restrict permissions on Network creation with ACL rules, or you can entirely disable the possibility to redefine the VLAN_ID by modifying the source code of /var/lib/one/remotes/vnm/ovswitch/OpenvSwitch.rb
. Change these lines:
if nic[:vlan_id] vlan = nic[:vlan_id] else vlan = CONF[:start_vlan] + nic[:network_id].to_i end
with this one:
vlan = CONF[:start_vlan] + nic[:network_id].to_i
To modify these rules you have to edit: /var/lib/one/remotes/vnm/ovswitch/OpenvSwitch.rb
.
Mac-spoofing
These rules prevent any traffic to come out of the port the MAC address has changed.
in_port=<PORT>,dl_src=<MAC>,priority=40000,actions=normal in_port=<PORT>,priority=39000,actions=normal
Black ports (one rule per port)
tcp,dl_dst=<MAC>,tp_dst=<PORT>,actions=drop
ICMP Drop
icmp,dl_dst=<MAC>,actions=drop