VRF on Linux using network namespaces

Question

My ultimate goal is to implement Virtual Routing and Forwarding (VRF) in Linux. The method that seems to be most widely accepted is to set up different network name spaces (one for each separate routing table) and run a Quagga or BIRD daemon for each namespace/routing table. I'm not married to this method, so if anyone has any other suggestions, please let me know.

The machine in question is running Debian 7 (wheezy) inside VMware workstation 12. It has always been a router and has been successfully routing for a while before I started this reconfiguration, so I know that the general routing setup is good.

The immediate problem is that I cannot communicate through my network namespace. That is, veth1 (which is in my namespace as below) can ping ONLY veth0 and nothing else. There is no network communication between veth1 and the networks below it - not even ARPs. If I didn't know better, I'd say someone had pulled the cable from the switch (but it's kind of hard to do that in virtual environment). And yes, I checked that the vmnets were properly set up. The router works when restored to its old configuration. It just doesn't work in this new configuration.

Anyone have any idea how to get veth1 communicating? Or even a wholly different method to get VRF working on Linux? Thanks in advance.

I set up the new configuration as follows:

add namespace

ip netns add nsx

add virtual interfaces

ip link add veth0 type veth peer name veth1

create a bridge

ip link add name vbr0 type bridge

add eth1 AND veth1 to the bridge

ip link set dev eth1 master vbr0
ip link set dev veth1 master vbr0

Assign veth1 to the namespace

ip link set veth1 netns nsx

configure the veth's IPs

ip addr add 10.0.2.10/24 dev vbr0
ip addr add 10.0.2.1/24 dev veth0
ip netns exec nsx ip addr add 10.0.2.2/24 dev veth1

Bring the i/f's up

ip link set dev vbr0 up
ip link set dev veth0 up
ip netns exec nsx ip link set dev veth1 up

assign veth2 its own routing table

ip netns exec nsx ip rule add dev veth1 table 1

Set the default route for the vtable

ip netns exec nsx ip route add default via 10.0.2.1 dev veth1

You can see in the output of iptables-save that everything is set to ACCEPT or FORWARD as appropriate

output of ip addr show:

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:10:e0:01 brd ff:ff:ff:ff:ff:ff
    inet 192.168.26.5/24 brd 192.168.26.255 scope global eth0
    inet 192.168.26.0/24 brd 192.168.26.255 scope global secondary eth0
    inet6 fe80::20c:29ff:fe10:e001/64 scope link
       valid_lft forever preferred_lft forever
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast master vbr0 state UP qlen 1000
    link/ether 00:0c:29:10:e0:ed brd ff:ff:ff:ff:ff:ff
    inet 10.0.0.1/24 brd 10.0.0.255 scope global eth1
    inet6 fe80::20c:29ff:fe10:e0ed/64 scope link
       valid_lft forever preferred_lft forever
47: veth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether ce:63:69:82:73:35 brd ff:ff:ff:ff:ff:ff
    inet 10.0.2.1/24 scope global veth0
    inet6 fe80::cc63:69ff:fe82:7335/64 scope link tentative
       valid_lft forever preferred_lft forever
48: vbr0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN
    link/ether 00:0c:29:10:e0:ed brd ff:ff:ff:ff:ff:ff
    inet 10.0.2.10/24 scope global vbr0
    inet6 fe80::20c:29ff:fe10:e0ed/64 scope link tentative
       valid_lft forever preferred_lft forever

output of ip route show:

 default via 192.168.26.2 dev eth0
    10.0.0.0/24 dev eth1  proto kernel  scope link  src 10.0.0.1
    10.0.2.0/24 dev vbr0  proto kernel  scope link  src 10.0.2.10
    10.0.2.0/24 dev veth0  proto kernel  scope link  src 10.0.2.1
    192.168.26.0/24 dev eth0  proto kernel  scope link  src 192.168.26.5

output of ip netns exec nsx ip addr show:

    45: lo: <LOOPBACK> mtu 16436 qdisc noop state DOWN
        link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    46: veth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
        link/ether b6:27:40:06:c2:de brd ff:ff:ff:ff:ff:ff
        inet 10.0.2.2/24 scope global veth1
        inet6 fe80::b427:40ff:fe06:c2de/64 scope link tentative
           valid_lft forever preferred_lft forever

output of ip netns exec nsx ip route show

 default via 10.0.2.1 dev veth1
    10.0.2.0/24 dev veth1  proto kernel  scope link  src 10.0.2.2

output of iptables-save

# Generated by iptables-save v1.4.14 on Thu Apr 14 18:19:19 2016
*nat
:PREROUTING ACCEPT [36:3588]
:INPUT ACCEPT [32:2540]
:OUTPUT ACCEPT [51:3744]
:POSTROUTING ACCEPT [55:4792]
COMMIT
# Completed on Thu Apr 14 18:19:19 2016
# Generated by iptables-save v1.4.14 on Thu Apr 14 18:19:19 2016
*filter
:INPUT ACCEPT [3319:373389]
:FORWARD ACCEPT [8:2004]
:OUTPUT ACCEPT [3558:428447]
COMMIT
# Completed on Thu Apr 14 18:19:19 2016

I implemented an utility which sets up network namespaces and all the required routes. Run it, see if it does what you expect it to do, then check out the source code. If you're unfamiliar with `node.js`, copy+paste one of the examples and run that. https://github.com/randunel/node-nns — randunel, Apr 17 '16 at 20:50
Looking at your code, what I think I see is: 1. I don't need a bridge. 2. I do need to enable ip forwarding separately in each namespace. Any comment? — user349251, Apr 18 '16 at 20:00
Sounds right. Make sure you loosen / remove filtering, including `arp` as well. — randunel, Apr 18 '16 at 21:18

score 2 · Accepted Answer · answered Aug 12 '16 at 00:00

2

Starting with kernel version 4.3 Linux has a VRF implementation. Checkout: https://www.kernel.org/doc/Documentation/networking/vrf.txt.

answered Aug 12 '16 at 00:00

DavidA

36
2

VRF on Linux using network namespaces

1 Answers1