netvirt-ha-docker

Run an ODL cluster on a single machine with these scripts.

Each ODL instance will run in its own docker container. Each container will mount its own overlay mount of your netvirt/karaf/target/assembly directory. This means that (a) all three containers share your single set of files comprising for e.g., the netvirt distribution, (b) the changes that each container makes to these files will not be propogated back to your assembly directory, (c) changes that each container makes are stored separately, on the host filesystem so that they can easily be inspected. Each container's files are stored ./mounts/[1|2|3]/assembly where [1|2|3] corresponds to the index of each container.

Three running (and empty) netvirt containers have been observed to occupy about 1.3GB of memory each.

Currently following projects are supported:

1. Netvirt
2. Genius
3. OVSDB Southbound
4. HWVTEP Southbound
5. Neutron

What scripts are provided

Script Name	Description
build	builds the overlay mounts for each container and the Docker image
up	configures the dedicated docker network used by the ODL containers (if it does not already exist) and launches the containers (but does not run ODL)
start	starts the ODL instances inside the containers
wait-for-active	Wait for the cluster to be active
wipe	completely deletes the docker containers
diagstatus	Convenience script to check whether an ODL is up and running using the diagstatus URL
shell	Convenience script to run commands or a shell on one of the containers
karaf-client	Convenience script to run karaf commands or a karaf shell on one of the containers

Understanding the Environment

These scripts will create and configure three containers:

Name	IP	Overlay Directory
odl_172.28.5.1	172.28.5.1	./mounts/1/assembly
odl_172.28.5.2	172.28.5.2	./mounts/2/assembly
odl_172.28.5.3	172.28.5.3	./mounts/3/assembly

The containers are networked together like this:

+----------------------------------------------------+
|                    HOST MACHINE                    |
|                                                    |
|    +------------+                                  |
|    | odl_       |                                  |
|    | 172.28.5.1 |----+                             |
|    +------------+    |                             |
|                      |                             |
|    +------------+    |                             |
|    | odl_       |----+--bridge: 172.28.5.254       |
|    | 172.28.5.2 |    |                             |
|    +------------+    |                             | 
|                      |                             |
|    +------------+    |                             |
|    | odl_       |----+                             |
|    | 172.28.5.1 |                                  |
|    +------------+                                  |
+----------------------------------------------------+

How to...

Before you start

1. You must have docker installed
2. pull the fedora image, docker pull fedora:latest
3. You must have your project cloned and project/karaf fully built (but not run)

Note that all scripts must be run from the root directory of this project (where this README file is).

Creating the containers

1. Build the overlay mounts and docker image: ./build <path to your project clone>. For OVSDB Southbound or HWVTEP the path should be '/southbound|hwvtepsouthbound'.

   N.B. that running this script will remove and erase any mount directories from previous runs. If you want to save logs, copy them aside.

2. Launch the docker containers: ./up

Running ODL

To start all ODLs running simply say:

./start

Then just wait for ./wait-for-active to exit.

However, since the initialization of ODL/karaf is extremely heavy it is best to just bring up two containers at first and only once they're up and running the third.

./start 1 2

Then wait for diagstatus to be ACTIVE (this can take a while):

./diagstatus 1

./diagstatus 2

N.B. that OVSDB projects currently don't support diagstatus, so you will have to check for OVSDB port [6640].

Only then should you

./start 3

Get into a container and debug stuff

You can enter the shell of a running container like this:

./shell 1 (or 2 or 3...you get the idea)

Run a command on a container

Execute a command on a running container:

./shell 1 ps ax

Run a command on all three containers

to execute a command on all three containers use the letter 'a' instead of a container index:

`./shell a find --name karaf.log

Open a karaf client on one of the containers

./karaf-client 1 (or 2 or 3)

Execute karaf commands on one of the containers

./karaf-client 1 feature:list

Execute a karaf command on all containers

`./karaf-client a log:set DEBUG org.opendaylight.genius

Forcefully remove the containers

To completely kill and wipe your running containers simply say "./wipe"