Amazon EC2 API with OpenStack - Developer Quick Start

Sat, Aug 9, 2014

OpenStack has support for EC2 API, that is, you can fire Amazon's API against an OpenStack cloud and it will still work. This article gets you started with using them locally against DevStack. It is more of a starter guide to a novice.

Fire a DevStack with it's default settings. See this post for more information on it.

git clone http://github.com/openstack-dev/devstack
cd devstack/
./stack.sh

Source openrc

source openrc

View all EC2 credentials available for the current user (here, demo user in demo tenant)

$ keystone ec2-credentials-list
+----------------------------------+----------------------------------+----------------------------------+
|              tenant              |              access              |              secret              |
+----------------------------------+----------------------------------+----------------------------------+
| 0e9f99a6f2064464aa054d305ba08052 | ef61007dae74468eb9593ffbbd22d9f1 | 28c7ad6248de4e6a8649b3e2d122ac5d |
| 9b93a67201264492be3d0998b87d821b | 1b0a617dbef347cb968c8eed160de0b3 | b6525738ad6044ea9c49abeefabf86de |
+----------------------------------+----------------------------------+----------------------------------+

But which one is my current tenant? Let's get that from parsing the output of token-get command

$ keystone token-get | grep tenant | awk '{print $4}'
0e9f99a6f2064464aa054d305ba08052

Note the access and secret keys.

Let's get started with the boto client for consuming AWS APIs. I prefer ipython shell, for its interactive features, but normal Python shell is just fine too. (Install ipython as sudo apt-get install ipython).

Import necessary module

>> import boto

Create a conn connection object, which we'll use for querying our cloud

>> conn = boto.connect_ec2_endpoint('http://10.0.1.126:8773/services/Cloud',
            aws_access_key_id='ef61007dae74468eb9593ffbbd22d9f1',
            aws_secret_access_key='28c7ad6248de4e6a8649b3e2d122ac5d')

Here 10.0.1.126 is the IP of my machine. Don't forget to change it to yours.

If everything is successful, call to get_all_instances() will return an empty list

>> conn.get_all_instances()
[]

Okay, now let's create an instance. List all the images first

In [20]: conn.get_all_images()
Out[20]:
[Image:aki-00000001,
 Image:ari-00000002,
 Image:ami-00000003,
 Image:ami-00000004]

Image ami-00000003 should be the cirros image from which we'll create an instance. But still, let's confirm that

In [26]: img = conn.get_image('ami-00000003')

In [27]: img.name
Out[27]: u'cirros-0.3.2-x86_64-uec'

Now let's use this image to create an instance. Boto's get_all_instances returns a list of reservations, which makes getting the instance object slightly roundabout.

In [35]: conn.run_instances(image_id='ami-00000003', instance_type='m1.tiny')
Out[35]: Reservation:r-08b8idoz

In [40]: reservations = conn.get_all_instances()

In [42]: resvn = reservations[0]

In [44]: instance = resvn.instances[0]

In [45]: instance.state
Out[45]: u'running'

And then delete it

In [47]: conn.terminate_instances('i-00000002')
Out[47]: [Instance:i-00000002]

In [50]: conn.get_all_reservations()
Out[50]: []

That's it for now :)

Use ipython or bpython for exploring boto library more and discover more APIs.

If you want to see what EC2 API was actually called behind the scenes, create a file /etc/boto.cfg and add these two lines. Now whenever you will use an interactive Python terminal, you'll see on your screen the EC2 API being called.

[Boto]
debug=2

Cheers!