openvswitch-formula

A saltstack formula for deploying OpenVSwitch.

Note

See the full Salt Formulas installation and usage instructions.

TODO

• remove network.managed from the openvswitch state as it's messing with /etc/network/interfaces
• get network.get_route(iface=None,dest=None) into SaltStack so we don't have to use a custom network-module
• add documentation for _modules/ovs_bridge, _states/ovs_bridge below
• add commandline examples to inline documentation (see the official bridge module's code)
• ovs_brigde.managed(force=True) to remove an interface when it's bonded to the wrong bridge?
• ovs_bridge.absent for an bridge that shouldn't exist?
• identify packages on different distributions and update map.jinja acordingly
• define a bool for kernel module via DKMS or not on Debian/Ubuntu
• make ovs_bridge.show() return a real dict not just txt that looks kinda like YAML (or JSON w/o curly braces)
• eventually move module from using cmdline tools to OVS' JSON RPC interface (probably based on code from OpenStack's Neutron)

Available states

• openvswitch
• networking.config

openvswitch

Installs the packages for openvswitch and starts the associated services though not distribution aware yet (probably won't work on RHEL/CentOS/etc.). Afterwards creates the bridges listed in pillar[openvswitch:bridges]. You can add list of interfaces to each bridge and those will be added as ports. If the parameter clean is set to True any other interfaces are removed. To re-use the configuration of a particular interface on the bridge set the parameter reuse_netcfg to the name of the interface.

openvswitch:
  bridges:
    - ovs-br0:
        ports:
          - eth0
        reuse_netcfg: eth0
    - ovs-br1:
        ports:
          - eth1
          - eth2
          - eth3
        clean: True

networking.config

Generates /etc/network/interfaces from pillar[interfaces], pillar[subnets] and pillar[openvswitch:bridges]. If an interface listed in pillar[interfaces] appears in a bridge's reuse_netcfg key its configuration will be used for the bridge. The interface itself will only get a minimal config to be set up and promisc on boot.

I.e. this pillar-data:

openvswitch:
  bridges:
    - ovs-br0:
        ports:
          - eth0
        reuse_netcfg: eth0
        comment: ovs-brigde for 10.10.0.0/16

interfaces:
  eth0:
    comment: Uplink for ovs-br0
    v4addr: 10.10.0.5/16

subnets:
  10.10.0.0/16:
    gateway: 10.10.0.1

Would result in something like this in your /etc/network/interfaces:

# ovs-brigde for 10.10.0.0/16
auto ovs-br0
iface ovs-br0 inet static
    address   10.10.0.5
    netmask       255.255.0.0
    network   10.10.0.0
    broadcast 10.10.255.255

# Uplink for ovs-br0
auto eth0
iface eth0 inet manual
    post-up ip link set promisc on eth0
    pre-down ip link set promisc off eth0

Available modules

• _modules/ovs_bridge
• _states/ovs_bridge

_modules/ovs_bridge

A module to manage OpenVSwitch bridges on supported Platforms (Linux and, in theory, FreeBSD and NetBSD). Its functions mirror those of the official bridge module.

_states/ovs_bridge

State module to get your OVS-bridges in the state you want them to be.