Skip to content

Latest commit

 

History

History
93 lines (75 loc) · 3.7 KB

README.md

File metadata and controls

93 lines (75 loc) · 3.7 KB
Raw

foreman-network-snippet

A snippet for foreman to do the networking, bonding, vlan tagging at build time using puppet + enc setup.

How it works

It leverages Puppet, expecting Puppet to be using Foreman as an ENC, in there you define the host data as parameters and such. When the box builds, it makes a query to its puppet master to get a dump of the parameters foreman has for the node and then it uses them to setup the networking.

It also reorganises the interfaces based on the BUS ID, this means, the first bus will become eth0, it will then map that busid to the mac of that device and create the interfaces in order from eth0 to eth based on the busid of the server.

Examples

Currently networking support in Foreman is very basic, it does not handle bonding / vlans / tags etc.

To get foreman to work with the networking i have had to write a snippet which will get executed in a %post, it fetches the data from foreman and then creates the necessary configuration files in /etc/sysconfig/network-scripts. The formant is:

device_name
device_address
device_gateway
device_netmask

Device can be any device (eth0, vlan123, bond0 etc.), but the principles mostly are the same throughout, except additional ones depending on the circumstances. A bond would need the interfaces associated with it as well as the bond options, so there will be additional:

# Standard interface no IP config
eth0_name = "Basic interface"

# Standard interface with IP config
eth0_name = "OOB eth0 interface"
eth0_address = "10.123.1.4"
eth0_netmask = "255.255.255.0"
eth0_gateway = "10.123.1.1"

# Bond interface with no IP config
bond0_interfaces = "eth0,eth3"
bond0_name       = "OOB bond"
bond0_options    = "mode=1 miimon=100 downdelay=200 updelay=200 primary=eth0 use_carrier=1"

# Bond interface with IP config
bond0_interfaces = "eth0,eth3"     # the interfaces associated with the bond i.e. eth0,eth3
bond0_address    = "10.123.3.4"    # the bonds IP if required
bond0_netmask    = "255.255.255.0" # the bond netmask if required (could be just a plain bond)
bond0_gateway    = "10.123.3.1"    # the gateway if required (may not be as just a plain bond)
bond0_options    = "mode=1 miimon=100 downdelay=200 updelay=200 primary=eth2 use_carrier=1 primary_reselect=2"                # the bond options
bond0_name       = "OOB BOND"      # the name of the bond (i.e. OOB)

# VLAN interface with no IP config
vlan1234_name  = "Multicast A Send"
vlan1234_device = "bond0"

# Vlan interface with IP configuration
vlan123_name     = "Multicast A send"    # the name of the vlan
vlan123_device   = "bond1"               # The device the vlan is associated with
vlan123_address  = "10.123.3.4"          # The IP of the interface (if required)
vlan123_netmask  = "255.255.255.0"       # the netmask of the interface (if required)
vlan123_gateway  = "10.123.3.1"          # the gateway aof the interface (if required)

The bond0 / vlan's and such can all be defined at different levels i.e. A group called CORE or BASE could contain:

bond0_interfaces = "eth0,eth3"
bond0_name = "OOB interface"
bond0_options = "mode=1 miimon=100"

I would then nest say "application x" inside CORE or BASE, which would inherit that bond setup, i could then add additional elements, OR override it if required. This would by standard, mean that all servers would get eth0,eth3 for bond0 and it would be a dedicated OOB, this would help standardise.

There is a snippet called network_setup, which is then included in the %post of the kickstart template

udev="/etc/udev/rules.d/70-persistent-net.rules"
[ -f $udev ] && rm -f $udev

%post --interpreter /usr/bin/python
<%= snippet "network_setup" %>