Difference between revisions of "VPP/Pure L3 Container Networking"

From fd.io
< VPP
Jump to: navigation, search
 
(3 intermediate revisions by the same user not shown)
Line 1: Line 1:
−
This example shows how to configure VPP as an IPv4 router interconnecting containers across multiple hosts.
+
This example shows how to configure VPP as an IPv4 router interconnecting containers across two hosts.
  
 
VPP itself runs in the root namespace, with a separate namespace for each container.
 
VPP itself runs in the root namespace, with a separate namespace for each container.
Line 5: Line 5:
 
[[File:Vpp-container.pdf|thumb|test network diagram]]
 
[[File:Vpp-container.pdf|thumb|test network diagram]]
  
−
== Setup ==
+
The basic setup is to use static ARPs and to unnumber the host interfaces from a loopback.  From that perspective this is a hybrid of Methods 1 and 2 from [[Pure_L3_Between_Namespaces_with_/32s]]
  
−
The 2 hosts are interconnected by a router.  The router has /24 routes for the client subnets - pointed at the appropriate vSwitch interfaces.
+
= Network Setup =
  
−
On each host:
+
The two hosts are interconnected by a router.  The router has /24 routes for the client subnets - pointed at the appropriate vSwitch interfaces.
  
−
<code>sudo docker create -e MICROSERVICE_LABEL=vpp -it --privileged -v "/tmp/vpp_socket:/tmp" -p 5001:5002 -p 9191:9191 --name vpp --network=host contivvpp/vswitch</code>
+
[[File:vpp-container.pdf]]
 +
 
 +
= Host Setup =
 +
 
 +
On each host do:
 +
 
 +
<code>
 +
sudo docker create -e MICROSERVICE_LABEL=vpp -it --privileged -v "/tmp/vpp_socket:/tmp" -p 5001:5002 -p 9191:9191 --name vpp --network=host contivvpp/vswitch
 +
</code>
 
   
 
   
 
create file vpp.conf as follows:
 
create file vpp.conf as follows:
  
 
<code>
 
<code>
−
unix {
+
unix {
−
:nodaemon
+
    nodaemon
−
:cli-listen 0.0.0.0:5002
+
    cli-listen 0.0.0.0:5002
−
:cli-no-pager
+
    cli-no-pager
−
}
+
}
−
dpdk {
+
dpdk {
−
:dev 0000:09:00.0 # replace this with an Ethernet interface on your host
+
    dev 0000:09:00.0 # replace this with an Ethernet interface on your host
−
:uio-driver igb_uio
+
    uio-driver igb_uio
−
}
+
}
 
</code>
 
</code>
  
 
then:
 
then:
−
 
 
<code>
 
<code>
−
sudo docker cp vpp.conf vpp:/etc/vpp/vpp.conf
+
sudo docker cp vpp.conf vpp:/etc/vpp/vpp.conf
−
<br />
+
sudo docker create -it --name client ubuntu  
−
sudo docker create -it --name client ubuntu  
+
 
</code>
 
</code>
  
 
(for the server host change the name to "server")
 
(for the server host change the name to "server")
 +
 +
== Client-Side Setup ==
 +
 +
from the Linux command line:
 +
 +
<code>
 +
sudo docker start vpp client
 +
export pid="$(sudo docker inspect -f '{{.State.Pid}}' "client")"
 +
sudo ln -sf /proc/$pid/ns/net /var/run/netns/client
 +
sudo ip link add name veth_client type veth peer name client
 +
sudo ip link set dev client up
 +
sudo ip link set dev veth_client up netns client
 +
export mac="$(sudo docker exec client ifconfig veth_client | awk 'NR==1{print $5}')"
 +
echo $mac
 +
export vmac="$(printf '
 +
set int ip address GigabitEthernet1/0/0 192.168.101.1/24\n
 +
set int state GigabitEthernet1/0/0 up\n
 +
create loopback interface\n
 +
set int ip address loop0 192.168.200.1/24\n
 +
set int state loop0 up\n
 +
create host-interface name client\n
 +
set int unnumbered host-client use loop0\n
 +
set ip arp host-client 192.168.200.2 MAC\n
 +
set int state host-client up\n
 +
ip route add 192.168.200.2/32 via 192.168.200.2 host-client\n
 +
ip route add 192.168.0.0/16 via 192.168.101.254 GigabitEthernet1/0/0\n
 +
show hardware-interfaces host-client\n
 +
quit' | sed s/MAC/$mac/ | nc 0 5002 | awk 'NR==29{print $3}')"
 +
echo $vmac
 +
sudo ip netns exec client ip link set dev lo up
 +
sudo ip netns exec client ip addr add 192.168.200.2/32 dev veth_client
 +
sudo ip netns exec client ip neigh add 192.168.200.1 lladdr $vmac dev veth_client
 +
sudo ip netns exec client ip route add 192.168.200.1 dev veth_client scope link
 +
sudo ip netns exec client ip route add 192.168.0.0/16 via 192.168.200.1 dev veth_client
 +
sudo ip netns exec client ip route add 1.2.3.4/32 via 192.168.200.1 dev veth_client
 +
sudo docker exec client ping -c 1 192.168.200.1
 +
</code>
 +
 +
the ping should succeed
 +
 +
== Server-Side Setup ==
 +
 +
Again from the Linux command line:
 +
 +
<code>
 +
sudo docker start vpp server
 +
export pid="$(sudo docker inspect -f '{{.State.Pid}}' "server")"
 +
sudo ln -sf /proc/$pid/ns/net /var/run/netns/server
 +
sudo ip link add name veth_server type veth peer name server
 +
sudo ip link set dev server up
 +
sudo ip link set dev veth_server up netns server
 +
export mac="$(sudo docker exec server ifconfig veth_server | awk 'NR==1{print $5}')"
 +
echo $mac
 +
export vmac="$(printf '
 +
set int ip address GigabitEthernet1/0/0 192.168.103.1/24\n
 +
set int state GigabitEthernet1/0/0 up\n
 +
create loopback interface\n
 +
set int ip address loop0 192.168.204.1/24\n
 +
set int state loop0 up\n
 +
create host-interface name server\n
 +
set int unnumbered host-server use loop0\n
 +
set ip arp host-server 192.168.204.2 MAC\n
 +
set int state host-server up\n
 +
ip route add 192.168.204.2/32 via 192.168.204.2 host-server\n
 +
ip route add 192.168.0.0/16 via 192.168.103.254 GigabitEthernet1/0/0\n
 +
show hardware-interfaces host-server\n
 +
quit' | sed s/MAC/$mac/ | nc 0 5002 | awk 'NR==29{print $3}')"
 +
echo $vmac
 +
sudo ip netns exec server ip link set dev lo up
 +
sudo ip netns exec server ip addr add 192.168.204.2/32 dev veth_server
 +
sudo ip netns exec server ip neigh add 192.168.204.1 lladdr $vmac dev veth_server
 +
sudo ip netns exec server ip route add 192.168.204.1 dev veth_server scope link
 +
sudo ip netns exec server ip route add 192.168.0.0/16 via 192.168.204.1 dev veth_server
 +
sudo docker exec server ping -c 1 192.168.204.1
 +
</code>
 +
 +
again the ping should succeed

Latest revision as of 15:51, 25 October 2017

This example shows how to configure VPP as an IPv4 router interconnecting containers across two hosts.

VPP itself runs in the root namespace, with a separate namespace for each container.

File:Vpp-container.pdf

The basic setup is to use static ARPs and to unnumber the host interfaces from a loopback. From that perspective this is a hybrid of Methods 1 and 2 from Pure_L3_Between_Namespaces_with_/32s

Network Setup

The two hosts are interconnected by a router. The router has /24 routes for the client subnets - pointed at the appropriate vSwitch interfaces.

File:Vpp-container.pdf

Host Setup

On each host do: 

sudo docker create -e MICROSERVICE_LABEL=vpp -it --privileged -v "/tmp/vpp_socket:/tmp" -p 5001:5002 -p 9191:9191 --name vpp --network=host contivvpp/vswitch

create file vpp.conf as follows: 

unix {
    nodaemon
    cli-listen 0.0.0.0:5002
    cli-no-pager
}
dpdk {
    dev 0000:09:00.0 # replace this with an Ethernet interface on your host
    uio-driver igb_uio
}

then: 

sudo docker cp vpp.conf vpp:/etc/vpp/vpp.conf
sudo docker create -it --name client ubuntu

(for the server host change the name to "server")

Client-Side Setup

from the Linux command line: 

sudo docker start vpp client
export pid="$(sudo docker inspect -f 'Template:.State.Pid' "client")"
sudo ln -sf /proc/$pid/ns/net /var/run/netns/client
sudo ip link add name veth_client type veth peer name client
sudo ip link set dev client up
sudo ip link set dev veth_client up netns client
export mac="$(sudo docker exec client ifconfig veth_client | awk 'NR==1{print $5}')"
echo $mac
export vmac="$(printf '
set int ip address GigabitEthernet1/0/0 192.168.101.1/24\n
set int state GigabitEthernet1/0/0 up\n 
create loopback interface\n
set int ip address loop0 192.168.200.1/24\n
set int state loop0 up\n
create host-interface name client\n
set int unnumbered host-client use loop0\n
set ip arp host-client 192.168.200.2 MAC\n
set int state host-client up\n
ip route add 192.168.200.2/32 via 192.168.200.2 host-client\n
ip route add 192.168.0.0/16 via 192.168.101.254 GigabitEthernet1/0/0\n
show hardware-interfaces host-client\n
quit' | sed s/MAC/$mac/ | nc 0 5002 | awk 'NR==29{print $3}')"
echo $vmac
sudo ip netns exec client ip link set dev lo up
sudo ip netns exec client ip addr add 192.168.200.2/32 dev veth_client
sudo ip netns exec client ip neigh add 192.168.200.1 lladdr $vmac dev veth_client
sudo ip netns exec client ip route add 192.168.200.1 dev veth_client scope link
sudo ip netns exec client ip route add 192.168.0.0/16 via 192.168.200.1 dev veth_client
sudo ip netns exec client ip route add 1.2.3.4/32 via 192.168.200.1 dev veth_client
sudo docker exec client ping -c 1 192.168.200.1

the ping should succeed

Server-Side Setup

Again from the Linux command line: 

sudo docker start vpp server
export pid="$(sudo docker inspect -f 'Template:.State.Pid' "server")"
sudo ln -sf /proc/$pid/ns/net /var/run/netns/server
sudo ip link add name veth_server type veth peer name server
sudo ip link set dev server up
sudo ip link set dev veth_server up netns server
export mac="$(sudo docker exec server ifconfig veth_server | awk 'NR==1{print $5}')"
echo $mac
export vmac="$(printf '
set int ip address GigabitEthernet1/0/0 192.168.103.1/24\n
set int state GigabitEthernet1/0/0 up\n
create loopback interface\n
set int ip address loop0 192.168.204.1/24\n
set int state loop0 up\n
create host-interface name server\n
set int unnumbered host-server use loop0\n
set ip arp host-server 192.168.204.2 MAC\n
set int state host-server up\n
ip route add 192.168.204.2/32 via 192.168.204.2 host-server\n
ip route add 192.168.0.0/16 via 192.168.103.254 GigabitEthernet1/0/0\n
show hardware-interfaces host-server\n
quit' | sed s/MAC/$mac/ | nc 0 5002 | awk 'NR==29{print $3}')"
echo $vmac
sudo ip netns exec server ip link set dev lo up
sudo ip netns exec server ip addr add 192.168.204.2/32 dev veth_server
sudo ip netns exec server ip neigh add 192.168.204.1 lladdr $vmac dev veth_server
sudo ip netns exec server ip route add 192.168.204.1 dev veth_server scope link
sudo ip netns exec server ip route add 192.168.0.0/16 via 192.168.204.1 dev veth_server
sudo docker exec server ping -c 1 192.168.204.1

again the ping should succeed

Retrieved from "https://wiki.fd.io/index.php?title=VPP/Pure_L3_Container_Networking&oldid=6065"