Lab: Taking the VPP Container Testbench for a Spin
Assuming the reader has already acquired the test bench build scripts, let’s start with building it.
1vagrant@ubuntu-focal$> make
2# Client image.
3DOCKER_BUILDKIT=1 docker build \
4 --file Dockerfile.vpp_testbench \
5 --build-arg HEALTHCHECK_PORT=8123 \
6 --tag vpp-testbench-client:local \
7 --target client_img \
8 .
9...
10...
11...
12DOCKER_BUILDKIT=1 docker build \
13 --file Dockerfile.vpp_testbench \
14 --build-arg HEALTHCHECK_PORT=8123 \
15 --tag vpp-testbench-server:local \
16 --target server_img \
17 .
18...
19...
20...
21 => exporting to image
22 => => exporting layers
23 => => writing image
24 => => naming to docker.io/library/vpp-testbench-server:local
250.0s
26Done.
Now, let’s start up our newly built images as a pair of containers. The various
hashes throughout this document will differ from those shown in your own
console (perfectly fine). First, let’s assume there are no running containers
on your system. We’ll verify via docker ps
:
1vagrant@ubuntu-focal$> docker ps
2CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
OK: good initial conditions. Now, let’s launch the containers:
1vagrant@ubuntu-focal$> make start
2# Create Docker bridge network.
3bash -c " . vpp_testbench_helpers.sh; host_only_create_docker_networks; "
46e071e533e239380b2fe92d6e0844c42736ec186226fbb20d89706f9a80f935f
5# Launch the containers.
6bash -c " . vpp_testbench_helpers.sh; host_only_run_testbench_client_container
7vpp-testbench-client:local; host_only_run_testbench_server_container
8vpp-testbench-server:local; "
9720fed0a94fd715694d73a41317f05a3f36860a6d5ae54db2d7cb7f2dcaf7924
10ccf166993d09e399f7b10d372c47cc9e72ce6092ef70ea206c699263da844e1b
11# Entrypoint scripts will bring up the various links.
12# Use "docker ps" to check status of containers, see if their health
13# probes are working as expected (i.e. "health"), etc.
Now, we can use docker ps
to verify if the containers are up and running:
1vagrant@ubuntu-focal$> docker ps
2CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
37d8e3ab35111 vpp-testbench-server:local "/entrypoint.sh" 3 seconds ago Up 2 seconds (health: starting) vpp-testbench-server
4cc01e64b12da vpp-testbench-client:local "/entrypoint.sh" 4 seconds ago Up 3 seconds (health: starting) vpp-testbench-client
Looking good so far. However, note the “health” status of the containers.
They’re not yet ready. We can re-execute the docker ps
command occasionally
until the containers are ready:
1vagrant@ubuntu-focal$> while true; do docker ps; sleep 1; done
2CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
342e9bcea7c58 vpp-testbench-server:local "/entrypoint.sh" 1 second ago Up Less than a second (health: starting) vpp-testbench-server
4710287b40bd3 vpp-testbench-client:local "/entrypoint.sh" 2 seconds ago Up Less than a second (health: starting) vpp-testbench-client
542e9bcea7c58 vpp-testbench-server:local "/entrypoint.sh" 30 seconds ago Up 29 seconds (health: starting) vpp-testbench-server
6710287b40bd3 vpp-testbench-client:local "/entrypoint.sh" 31 seconds ago Up 30 seconds (healthy) vpp-testbench-client
7CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
842e9bcea7c58 vpp-testbench-server:local "/entrypoint.sh" 31 seconds ago Up 30 seconds (healthy) vpp-testbench-server
9710287b40bd3 vpp-testbench-client:local "/entrypoint.sh" 32 seconds ago Up 31 seconds (healthy) vpp-testbench-client
Not the most elegant approach, but it works. Onward.
Note
How would one automate this step so that we’re not having to manually watch
the console until the containers are ready? What’s something that we could
put into a script our our Makefile
to poll the containers until they’re
ready to use?
Spoiler
# "Direct" approach.
while true; do
[ '"healthy"' = docker inspect --format "{{json .State.Health.Status }}" vpp-testbench-client] && break
done
# Could also use awk/grep/etc. against the output of "docker ps".
Now that our containers are up and running, let’s drop a shell into the “client” container:
1vagrant@ubuntu-focal$> make shell_client
First, let’s take a look at the default network configuration.
1root@478ab126035e:/work# ip a
21: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
3 link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
4 inet 127.0.0.1/8 scope host lo
5 valid_lft forever preferred_lft forever
62: vxlan-vid-42: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UNKNOWN group default qlen 1000
7 link/ether 3a:2c:19:cb:ca:35 brd ff:ff:ff:ff:ff:ff
8 inet 169.254.10.1/24 scope global vxlan-vid-42
9 valid_lft forever preferred_lft forever
103: vpp-tap-0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UNKNOWN group default qlen 1000
11 link/ether 02:fe:c5:52:63:12 brd ff:ff:ff:ff:ff:ff
12 inet 169.254.12.1/24 scope global vpp-tap-0
13 valid_lft forever preferred_lft forever
14635: eth0@if636: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default
15 link/ether 02:42:a9:fe:00:01 brd ff:ff:ff:ff:ff:ff link-netnsid 0
16 inet 169.254.0.1/24 brd 169.254.0.255 scope global eth0
17 valid_lft forever preferred_lft forever
Let’s also enumerate the interfaces managed by VPP. For the help of the reader,
there is a shell function, vc
, which just launches vppctl
with some
helpful default arguments.
1root@478ab126035e:/work# type vc
2vc is a function
3vc ()
4{
5 vppctl -s "${VPP_SOCK}" "${@}"
6}
7root@478ab126035e:/work# vc
8 _______ _ _ _____ ___
9 __/ __/ _ \ (_)__ | | / / _ \/ _ \
10 _/ _// // / / / _ \ | |/ / ___/ ___/
11 /_/ /____(_)_/\___/ |___/_/ /_/
12
13vpp# show int
14 Name Idx State MTU (L3/IP4/IP6/MPLS) Counter Count
15local0 0 down 0/0/0/0
16memif0/0 1 up 9000/0/0/0 rx packets 7
17 rx bytes 521
18 tx packets 7
19 tx bytes 503
20tap0 2 up 9000/0/0/0 rx packets 7
21 rx bytes 503
22 tx packets 7
23 tx bytes 521
24vpp#
Note
One more exercise for the reader:
1. From the client container, how would you ping the server container on the
Linux-managed VXLAN interface?
2. From the client container, how would you ping the server container on the
VPP-managed TAP interface?
3. A couple trivial web servers (using netcat
) are running on the server
container. Besides looking at the Makefile
recipes, how could one
determine what ports and interfaces these servers are bound to, and
how would one issue an HTTP GET query against them from the client
container? (hint: you’re allowed to log-in to the server container via
make shell_server
, and the netstat
command may be of use).
Spoiler
1. ping 169.254.10.2
2. ping 169.254.12.2
3. make shell_server
netstat -tulpn
tcp 0 0 169.254.12.2:8000 0.0.0.0:*
LISTEN 47/nc
tcp 0 0 169.254.10.2:8000 0.0.0.0:*
LISTEN 34/nc
exit
make shell_client
root@478ab126035e:/work# curl 169.254.10.2:8000
HOST:14f0df855445
DATE:Fri Nov 19 16:36:57 UTC 2021
Hello from the Linux interface.
root@478ab126035e:/work# curl 169.254.12.2:8000
HOST:14f0df855445
DATE:Fri Nov 19 16:37:04 UTC 2021
Hello from the VPP interface.
exit
Now that we’ve done some quick exercises, let’s clean-up the containers and their associated resources.
1vagrant@ubuntu-focal$> make stop
2# Terminate the containers.
3bash -c " . vpp_testbench_helpers.sh; host_only_kill_testbench_client_container vpp-testbench-client:local; host_only_kill_testbench_server_container vpp-testbench-server:local; "
4vpp-testbench-client
5Error: No such container: vpp-testbench-client
6vpp-testbench-server
7Error: No such container: vpp-testbench-server
8# Cleanup Docker bridge network.
9bash -c " . vpp_testbench_helpers.sh; host_only_destroy_docker_networks; "
10vpp-testbench-net
That’s it for this section.