The network statement as shown works properly if encoded either in the Network field of Advanced Options or if placed inline in the Partitions section. If used, this way, 3 interfaces are created, bond0 and it's 2 components eth0 and eth1. The LACP bond apparently works as one would expect.

The problem is if I generate this exact statement via an echo in a snippet, or via writes to a temporary file which is included via a %include in the Partitions section. I've seen both methods described elsewhere but can't make them work. If I use the %include method, the resulting file is automatically copied to root's home directory. It is correct in this file. If I use the echo method, The output of the echo is listed in the log file for the snippet.

From my little experience, apparently everything works with the host side not running LACP and the switch side running LACP. That's not as I expected. I expected Anaconda to fail altogether. I'm just saying that you have to skip through the apparent Anaconda errors and then everything will progress normally.

I realize that I could do everything with a post-script, but the various blogs seem to say that it can do network config with pre-scripts and %includes. I'd like to learn how to crack that nut.

Re: %include /tmp/netinfo

RedHat documents indicate that this method is no longer supported as NetworkManager is used. Therefore, the /etc/sysconfig/network-scripts must be populated directly. I'll have to give this a try.

RE: bonding

I did some further testing. Bonding can be configured either by using the network field on Advanced Options or by unchecking the Network box in Advanced Options and placing the entire network command in Partitioning. If you place it in Partitioning, but leave the Network box checked in Advanced Options, the bonding configuration will be ignored.

Well, I'm not sure I'd say bonding works during kickstart. What I'd say is I don't see a network failure if LACP Mode 4 is defined on the switch and you try to use one of the interfaces to do an install. The reason I say it this way is what I see when I try to do the install.

As I said previously, when you do the install, Anaconda gives an error saying that it can't load something from the Satellite server. I don't remember what it is. If you then select OK, Anaconda will continue to a screen where you select which network adapter you wish to install from. The bonded adapter is not listed here, just eth0 and eth1. I select eth0 and the install then completes normally. This leads me to believe that Anaconda is only using one interface even thought the switch is set for LACP.

This kind of makes sense from what the network guru told me. He said that the way LACP Mode 4 works is that the switch just listens on all ports defined for the LACP port group. Whichever one responds, the switch associates that port with the MAC address it receives on that port. When the switch needs to send a packet for that MAC address, it just checks which port last sent data with that MAC address and sends data on that port. If the server sends data on a different port, it just reassociates the MAC address with this new port.

If this descripttion is accurate, it doesn't matter if the host is actually using a bonded interface as long as it only uses 1 interface and only sends 1 MAC address on that 1 interface. There's no confusion on the switch side because it will associate the MAC address to which ever port used to receive the last packet. The server doesn't care because it will only get packets on the one interface and he will get packets with the correct MAC address.

Technically, this kind of configuration may not be supported. It seems to work, though.

I simply wrote a snippet that updated the files in /etc/sysconfig/network-scripts/ifcfg_*

O.K. Here are some Python snippets. The first is the actual config data. It's just Python dictionaries (or hashes). Should be fairly obvious. This is where you put your network configuration. The second is the actual code snippet. I'm a fairly new Python user (I learned exclusively for Satellite) so I'm sure this could be coded a bit better.

Define your script and put the snippet call in it:
$SNIPPET('spacewalk/1/base-post-hw-network')

Name your script ( call mine the same as the snippet:
base-post-hw-network

Define the scripting language:
/usr/bin/python

Select Script Execution Time:
Post Script

Check "Template"

Click Update

--------------@config-network----------------
DNS=["10.10.10.36", "10.10.10.17", "10.10.10.38"];
DOMAIN="ccci.org"
NAMESERVERS="ccci.org,net.ccci.org"
VLANS={
10:{
'NETMASK':'255.255.255.0',
'GATEWAY':'10.10.10.254'
},
11:{
'NETMASK':'255.255.255.0',
'GATEWAY':'10.10.11.254'
},
13:{
'NETMASK':'255.255.255.0',
'GATEWAY':'10.10.13.254'
},
417:{
'NETMASK':'255.255.255.192',
'GATEWAY':'10.10.210.1'
}
}
IFACES={
"00:18:8B:32:84:67":
{
'HOSTNAME':"llrhnt00",
'IPADDR':"10.10.11.9",
'MASTER':"bond0",
'VLAN':11
},
"00:18:8B:32:84:69":
{
'HOSTNAME':"llrhnt00",
'IPADDR':"10.10.11.9",
'MASTER':"bond0",
'VLAN':11
},
"B8:CA:3A:6E:9D:C0":
{
'HOSTNAME':"plorad60",
'IPADDR':"10.10.210.60",
'MASTER':"bond0",
'VLAN':417
},
"B8:CA:3A:6E:9D:C1":
{
'HOSTNAME':"plorad60",
'IPADDR':"10.10.210.60",
'MASTER':"bond0",
'VLAN':417
},
}

--------------base-post-hw-network----------------
import re
import platform
import subprocess
import glob
import uuid
from subprocess import Popen

$SNIPPET('spacewalk/1/@config-network')

def run(cmd):
popen=Popen(cmd, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
return popen.stdout

def update_iface(from_iface, to_iface):
for k, v in from_iface.iteritems():
to_iface[k] = v

def update_ifaces(from_ifaces, to_ifaces):
for from_key, from_iface in from_ifaces.iteritems():
if from_key in to_ifaces:
to_iface=to_ifaces[from_key]
update_iface(from_ifaces[from_key], to_iface)
to_ifaces[from_key]=to_iface
else:
to_ifaces[from_key]=from_iface

p1=re.compile('(?P[^=]+)="?(?P[^"]+)"?')
p2=re.compile(r"^(?P\w+) +Link +encap:Ethernet +HWaddr +(?P[0-9[a-fA-F:]+).*")

HOSTNAME=platform.node()
DOMAIN="ccci.org"
SEARCH="ccci.org,net.ccci.org"
DNS=["10.10.10.36", "10.10.10.17", "10.10.10.38"];
DNSNAMES=""
for n in range(0, len(DNS)):
DNSNAMES+="DNS{0}={1}\n".format(n+1, DNS[n])

$SNIPPET('spacewalk/1/@config-network')

uppercase keys in IFACES, user might have forgotten

ifaces={}
for k1, v1 in IFACES.iteritems():
iface={}
for k2, v2 in v1.iteritems():
iface[k2.upper()]=v2
ifaces[k1.upper()]=v1
IFACES=ifaces

apply available device names from ifconfig to IFACES

p=run("ifconfig -a")
for line in p.readlines():
m2=p2.match(line)
if m2:
DEVICE=m2.group('DEVICE')
HWADDR=m2.group('HWADDR')
if HWADDR in IFACES:
iface=IFACES[HWADDR]
iface['DEVICE']=DEVICE
iface['HOSTNAME']=HOSTNAME
iface['HWADDR']=HWADDR
IFACES[HWADDR]=iface

build new ifaces from ifcfg files

ifaces={}
for file in glob.glob('/etc/sysconfig/network-scripts/ifcfg-*'):
if file == '/etc/sysconfig/network-scripts/ifcfg-lo':
continue
iface={}
f=open(file, 'r')
for line in f.readlines():
line=line.strip()
m1=p1.match(line)
if m1:
iface[m1.group('name')]=m1.group('value')
f.close()
if 'HWADDR' in iface:
HWADDR=iface['HWADDR']
ifaces[HWADDR]=iface

merge new ifaces with original IFACES

update_ifaces(ifaces, IFACES)

build list of bonded interfaces from interfaces

and fill in device names from ifconfig

dev=""
bonds={}
nets={}
for HWADDR, IFACE in IFACES.iteritems():
if 'DEVICE' in IFACE:
DEVICE=IFACE['DEVICE']
nets[IFACE['DEVICE']]=IFACE
if 'MASTER' in IFACE:
MASTER=IFACE['MASTER']
IFACE['SLAVE']='yes'
bond={}
if MASTER in bonds:
bond=bonds[MASTER]
if not 'HOSTNAME' in IFACE:
print IFACE
exit(0)
bond[DEVICE]=IFACE
bonds[MASTER]=bond
ifaces[HWADDR]=IFACE
for hwaddr, iface in ifaces.iteritems():
IFACES[HWADDR]=iface

emit any defined bond interfaces

for bondname, bond in bonds.iteritems():
bonddevs=list(bond.keys())
bondkey0=bonddevs[0]
bondx=bond[bondkey0]
bondname=bond[bondkey0]['MASTER']
HOSTNAME=bond[bondkey0]['HOSTNAME']
IPADDR=bond[bondkey0]['IPADDR']
VLAN=VLANS[bond[bondkey0]['VLAN']]
GATEWAY=VLAN['GATEWAY']
NETMASK=VLAN['NETMASK']
UUID=uuid.uuid4()
bondnames=",".join(bonddevs)
t=open('/etc/sysconfig/network-scripts/ifcfg-{0}'.format(bondname), 'w')
t.write("#-----\nDEVICE={0}\nHOSTNAME={1}.{2}\nIPADDR={3}\nGATEWAY={4}\nNETMASK={5}\n{6}UUID={7}\nIPV6INIT=yes\nMTU=1500\nNM_CONTROLLED=no\nONBOOT=yes\nTYPE=bond\nBONDING_OPTS=\"mode=802.3ad miimon=100 lacp_rate=fast\"\nBOOTPROTO=static\n".format(bondname,HOSTNAME,DOMAIN,IPADDR,GATEWAY,NETMASK,DNSNAMES,UUID))
t.close()
t=open('/etc/modprobe.d/{0}.conf'.format(bondname), 'w')
t.write('alias {0} bonding\n'.format(bondname))
t.close()

emit real interfaces

for netname, net in nets.iteritems():
if not 'HOSTNAME' in net:
continue
SLAVE=''
if 'SLAVE' in net:
SLAVE=net['SLAVE'].upper()
del net['VLAN']
if SLAVE == 'YES':
net['ONBOOT']='yes'
net['NM_CONTROLLED']='no'
net['BOOTPROTO']='none'
for k in ['GATEWAY', 'NETMASK', 'IPADDR', 'TYPE', 'HOSTNAME']:
if k in net:
del net[k]
t=open("/etc/sysconfig/network-scripts/ifcfg-{0}".format(netname), 'w')
t.write('#---------\n')
for k, v in net.iteritems():
k=k.upper()
# don't emit stuff on slave interface
t.write('{0}="{1}"\n'.format(k, v))
t.close()