koko is a simple tool which connects between Docker containers/linux netns processes with veth devices/vxlan of linux kernel. Using koko, you can simply make point-to-point connection for containers without linux bridges.

koko supports following container:

• Docker
• Linux netns namespace (i.e. 'ip netns' or see 'man ip-netns' for its detail)

koko supports following linux interface to connect above:

• veth: Virtual Ethernet Interface
• vxlan: Virtual eXtensible Local Area Network

See releases page.

• golang 1.7 or later

koko is written in go, so following commands makes koko single binary. Build and put it in your container host.

go get github.com/redhat-nfvpe/koko 

koko takes two arguments: two endpoints of container and koko connects both. koko supports veth for two containers in one host and vxlan for two containers in separate host.

./koko {-c <linkname> | -d <container>,<linkname>[,<IP addr>/<prefixlen>,...] | -n <netns name>,<linkname>[,<IP addr>/<prefixlen>,...]| -p <pid>,<linkname>[,<IP addr>/<prefixlen>,...]| -c <linkname> } {-d <container>,<linkname>[,<IP addr>/<prefixlen>,...] | -n <netns name>,<linkname>[,<IP addr>/<prefixlen>,...]| -p <pid>,<linkname>[,<IP addr>/<prefixlen>,...]| -c <linkname> } 

Connecting containers which are in separate hosts with vxlan. Following command makes vxlan interface and put this interface into given container with/without IP address.

./koko {-c <linkname> | -d <container>,<linkname>[,<IP/mirror>,...] | -n <netns name>,<linkname>[,<IP/mirror>,...]| -p <pid>,<linkname>[,<IP/mirror>,...] } -x <parent interface>,<remote endpoint IP addr>,<vxlan id> <IP/mirror> = {<IP addr>/<prefixlen> | mirror:{ingress|egress|both},<mirror IF>} 

Connecting containers which are in separate hosts with vlan. Following command makes vlan interface and put this interface into given container with/without IP address.

./koko {-c <linkname> | -d <container>,<linkname>[,<IP/mirror>,...] | -n <netns name>,<linkname>[,<IP/mirror>,...]| -p <pid>,<linkname>[,<IP/mirror>,...] } -V <parent interface>,<vlan id> <IP/mirror> = {<IP addr>/<prefixlen> | mirror:{ingress|egress|both},<mirror IF>} 

Connecting containers which are in separate hosts with macvlan. Following command makes macvlan interface and put this interface into given container with/without IP address.

./koko {-c <linkname> | -d <container>,<linkname>[,<IP/mirror>,...] | -n <netns name>,<linkname>[,<IP/mirror>,...]| -p <pid>,<linkname>[,<IP/mirror>,...] } -M <parent interface>,<macvlan mode, {default|private|vepa|bridge|passthru}> <IP/mirror> = {<IP addr>/<prefixlen> | mirror:{ingress|egress|both},<mirror IF>} 

koko -D and koko -N deletes veth interface or vxlan interface. In case of veth, peering interface is also removed in this command.

./koko {-D <container>,<linkname> | -N <netns name>,<linkname> } 

In case of 'egress' (and 'both'), the target interface (i.e. ) needs to be configured to have a queue because veth does not have tx queue in default (see moby/moby#33162 for the details). ip link set <mirror IF> qlen <queue length> sets queue length to corresponding veth device.

• -c is to create veth and put it in current namespace
• -C is to delete veth in current namespace
• -d is to create interface and put it in docker container namespace
• -D is to delete interface of docker container namespace
• -e is to create interface and put it in cri-o container namespace
• -E is to delete interface of cri-o container namespace
• -n is to create interface and put it in linux netns namespace
• -N is to delete interface of linux netns namespace
• -p is to create interface and put it in pid's netns namespace
• -P is to delete interface of pid's netns namespace
• -X is to create vxlan interface
• -V is to create vlan interface
• -M is to create macvlan interface
• -h is to show help
• -v is to show version

./koko -h 

Please see Examples in Wiki.

# connect between docker containers with IPv4 address sudo ./koko -d centos1,link1,192.168.1.1/24 -d centos2,link2,192.168.1.2/24 # connect between docker containers with IPv6 address sudo ./koko -d centos1,link1,2001:DB8::1/64 -d centos2,link2,2001:DB8::2/64 # connect between docker containers with IPv4/IPv6 address sudo ./koko -d centos1,link1,192.168.1.1/24,2001:DB8::1/64 -d centos2,link2,192.168.1.2/24,2001:DB8::2/64 # connect between netns namespaces sudo ./koko -n testns1,link1,192.168.1.1/24 -n testns2,link2,192.168.1.2/24 # connect between docker container and netns namespace sudo ./koko -d centos1,link1,192.168.1.1/24 -n testns2,link2,192.168.1.2/24 # create vxlan interface and put it into docker container sudo ./koko -d centos1,link1,192.168.1.1/24 -x eth1,10.1.1.1,1 

• Document

• Tomofumi Hayashi (s1061123)
• Doug Smith (dougbtv)

• Antonio Ojea (aojea)