bms-network-setup

Script to parse network_data.json and configure ifcfg-* files from it.

How it works

This is a limited and simple script to enable networking on OTC Bare Metal instances. Bare Metal instances have a ConfigDrive that contains meta_data.json and network_data.json. The meta_data.json is unfortunately not fully correct at this point (wrong hostname), so we prefer to use the network OpenStack DataSource via the normal cloud-init mechanism. However, to even get networking working, we need to parse and evaluate network_data.json. This scripts does this by parsing it and then writing out the network ifcfg-* files and then leaves it to the standard distribution networking mechanisms (wicked, ifup, NetworkManager, ...) to do the network setup and leave it to cloud-init for all other cloud-init steps.

In our package (on OpenBuildService in home:garloff:OTC), we also have the mechanism to dynamically disable the ConfigDrive data source IF running on an OTC BMS.

The script can produce network interface config files for modern SUSE, RedHat and Debian based Linux distributions.

Parameters

Usage: bms-network-setup.py [-d] [-s|u|r|n]
 -d: Debug: reads network_data.json and writes ifcfg-* in current dir
 -s: SuSE: assume we run on a SuSE distribution
 -u: Debian: assume we run on a Debian/Ubuntu distribution
 -r: RedHat: assume we run on a RedHat/CentOS distribution
 -n: Netplan: assume we run on a distribution using netplan

By default, the script produces ifcfg files for the distribution it detects it's currently running on.

Alternatives

This is a different approach from bms-network-config. bms-network-config does a lot more and duplicates some of the things that cloud-init and the distro network setup scripts/NM/wickedd do.

cloud-init does have support to parse network_data.json -- just not complete enough to support Bare Metal instances in OTC. Medium term, having cloud-init enhanced to do the job would be the preferred option.

Examples

network_data.json

{
  "services": [
    {
      "type": "dns",
      "address": "100.125.4.25"
    }, {
      "type": "dns",
      "address": "8.8.8.8"
    } ],
  "networks": [
    {
      "network_id": "fc282960-6dd3-4797-8e6d-33389de9671d",
      "link": "interface0",
      "type": "ipv4_dhcp",
      "id": "network0",
      "gateway": "192.168.66.1"
    }, {
      "network_id": "a90ad03f-4cb4-4cb3-8b10-a8aa571a169f",
      "link": "interface1",
      "type": "ipv4_dhcp",
      "id": "network1",
      "gateway": "192.168.70.1"
    }, {
      "network_id": "fbddc7a6-6d2b-4957-a880-aa89238ab011",
      "type": "ipv4",
      "netmask": "255.255.252.0",
      "link": "interface2",
      "routes": [],
      "ip_address": "192.168.81.99",
      "id": "network2"
    } ],
  "links": [
    {
      "type": "phy",
      "ethernet_mac_address": "2c:55:d3:9a:06:35",
      "id": "eth0",
      "name": "eth0",
      "mtu": 8888
    }, {
      "type": "phy",
      "ethernet_mac_address": "2c:55:d3:9a:06:36",
      "id": "eth1",
      "name": "eth1",
      "mtu": 8888
    }, {
      "bond_miimon": 100,
      "ethernet_mac_address": "fa:16:3e:3f:15:55",
      "mtu": 8888,
      "bond_mode": "1",
      "bond_links": [ "eth0", "eth1" ],
      "type": "bond",
      "id": "interface0"
    }, {
      "ethernet_mac_address": "fa:16:3e:87:0b:67",
      "mtu": 8888,
      "vlan_link": "interface0",
      "vlan_id": 2315,
      "type": "vlan",
      "id": "interface1",
      "vlan_mac_address": "fa:16:3e:87:0b:67"
    }, {
      "ethernet_mac_address": "fa:16:3e:fc:a9:28",
      "mtu": 8888,
      "vlan_link": "interface0",
      "vlan_id": 3966,
      "type": "vlan",
      "id": "interface2",
      "vlan_mac_address": "fa:16:3e:fc:a9:28"
    } ]
}

bms-network-setup.py creates the ifcfg-eth0, -eth1, -bond0, -vlan2315, and -vlan3996 files. bond0 will enslave eth0 and eth1, creating an bond link (mode1 => active/passive). vlan2315 and vlan3996 are VLANs on this bond0 device, the former configured with DHCP while the latter has a static configuration.