Now that we’ve got that covered, let’s look deeper into a very cool feature – the ability to customize the instance and automate its startup and configuration. For example, at instance creation time you can specify a snappy application to be installed. cloud-init is what allows you to do this, and it is installed inside the Snappy image. cloud-init receives this information from the user in the form of ‘user-data’.
One of the formats that can be fed to cloud-init is called ‘cloud-config’. cloud-config is yaml formatted data that is interpreted and acted on by cloud-init. For Snappy, we’ve added a couple specific configuration values. Those are included under the top level ‘snappy’.
- ssh_enabled: determines if ‘ssh’ service is started or not. By default ssh is not enabled.
- packages: A list of snappy packages to install on first boot. Items in this list are snappy package names.
When running inside snappy, cloud-init still provides many of the features it provides on traditional instances. Some useful configuration entries:
- runcmd: A list of commands run after boot has been completed. Commands are run as root. Each entry in the list can be a string or a list. If the entry is a string, it is interpreted by ‘sh’. If it is a list, it is executed as a command and arguments without shell interpretation.
- ssh_authorized_keys: This is a list of strings. Each key present will be put into the default user’s ssh authorized keys file. Note that ssh authorized keys are also accepted via the cloud’s metadata service.
- write_files: this allows you to write content to the filesystem. The module is still expected to work, but the user will have to be aware that much of the filesystem is read-only. Specifically, writing to file system locations that are not writable is expected to fail.
Some cloud-init config modules are simply not going to work. For example, traditional packages will not be installed by ‘apt’ as the root filesystem is read-only.
Example Cloud Config
Its always easiest to start from a working example. Below is one that demonstrates the usage of the config options listed above. Please note that user data intended to be consumed as cloud-config must contain the first line ‘#cloud-config‘.
– content: |
echo “==== Hello Snappy! It is now $(date -R) ====”
– /writable/greet | tee /run/hello.log
Launching with uvtool
Follow yesterday’s blog post to get a functional tool. Then, save the example config file above to a file, and launch you’re instance with it.
$ uvt-kvm create –wait –add-user-data=my-config.yaml snappy1 release=devel
Our user-data instructed cloud-init to do a number of different things. First, it wrote a file via ‘write_files’ to a writable space on disk, and then executed that file with ‘runcmd’. Lets verify that was done:
$ uvt-kvm ssh snappy1 cat /run/hello.log
==== Hello Snappy! It is now Thu, 11 Dec 2014 18:16:34 +0000 ====
It also instructed cloud-init to install the Snappy ‘xkcd-webserver’ application.
$ uvt-kvm ssh snappy1 snappy versions
Part Tag Installed Available Fingerprint Active
ubuntu-core edge 141 – 7f068cb4fa876c *
xkcd-webserver edge 0.3.1 – 3a9152b8bff494 *
There we can see that xkcd-webserver was installed, lets check that it is running:
$ uvt-kvm ip snappy1
$ wget -O – –quiet http://192.168.122.80/ | grep <title>
Launching on Azure
The same user-data listed above also works on Microsoft Azure. Follow the instructions for setting up the azure command line tools, and then launch the instance with and provide the ‘–custom-data‘ flag. A full command line might look like:
$ azure vm create snappy-test $imgid ubuntu \
–location “North Europe” –no-ssh-password \
–ssh-cert ~/.ssh/azure_pub.pem –ssh \
Have fun playing with cloud-init!