What is a datasource?

Datasources are sources of configuration data for cloud-init that typically come from the user (aka userdata) or come from the stack that created the configuration drive (aka metadata). Typical userdata would include files, yaml, and shell scripts while typical metadata would include server name, instance id, display name and other cloud specific details. Since there are multiple ways to provide this data (each cloud solution seems to prefer its own way) internally a datasource abstract class was created to allow for a single way to access the different cloud systems methods to provide this data through the typical usage of subclasses.

The current interface that a datasource object must provide is the following:

# returns a mime multipart message that contains
# all the various fully-expanded components that
# were found from processing the raw userdata string
# - when filtering only the mime messages targeting
#   this instance id will be returned (or messages with
#   no instance id)
def get_userdata(self, apply_filter=False)

# returns the raw userdata string (or none)
def get_userdata_raw(self)

# returns a integer (or none) which can be used to identify
# this instance in a group of instances which are typically
# created from a single command, thus allowing programatic
# filtering on this launch index (or other selective actions)
def launch_index(self)

# the data sources' config_obj is a cloud-config formated
# object that came to it from ways other than cloud-config
# because cloud-config content would be handled elsewhere
def get_config_obj(self)

#returns a list of public ssh keys
def get_public_ssh_keys(self)

# translates a device 'short' name into the actual physical device
# fully qualified name (or none if said physical device is not attached
# or does not exist)
def device_name_to_device(self, name)

# gets the locale string this instance should be applying
# which typically used to adjust the instances locale settings files
def get_locale(self)

def availability_zone(self)

# gets the instance id that was assigned to this instance by the
# cloud provider or when said instance id does not exist in the backing
# metadata this will return 'iid-datasource'
def get_instance_id(self)

# gets the fully qualified domain name that this host should  be using
# when configuring network or hostname releated settings, typically
# assigned either by the cloud provider or the user creating the vm
def get_hostname(self, fqdn=False)

def get_package_mirror_info(self)


The EC2 datasource is the oldest and most widely used datasource that cloud-init supports. This datasource interacts with a magic ip that is provided to the instance by the cloud provider. Typically this ip is of which at this ip a http server is provided to the instance so that the instance can make calls to get instance userdata and instance metadata.

Metadata is accessible via the following URL:


Userdata is accessible via the following URL:

1234,fred,reboot,true | 4512,jimbo, | 173,,,

Note that there are multiple versions of this data provided, cloud-init by default uses 2009-04-04 but newer versions can be supported with relative ease (newer versions have more data exposed, while maintaining backward compatibility with the previous versions).

To see which versions are supported from your cloud provider use the following URL:


Config Drive

The configuration drive datasource supports the OpenStack configuration drive disk.

See the config drive extension and introduction in the public documentation for more information.

By default, cloud-init does always consider this source to be a full-fledged datasource. Instead, the typical behavior is to assume it is really only present to provide networking information. Cloud-init will copy off the network information, apply it to the system, and then continue on. The “full” datasource could then be found in the EC2 metadata service. If this is not the case then the files contained on the located drive must provide equivalents to what the EC2 metadata service would provide (which is typical of the version 2 support listed below)

Version 1

The following criteria are required to as a config drive:

  1. Must be formatted with vfat filesystem
  2. Must be a un-partitioned block device (/dev/vdb, not /dev/vdb1)
  3. Must contain one of the following files


This file is laid down by nova in order to pass static networking information to the guest. Cloud-init will copy it off of the config-drive and into /etc/network/interfaces (or convert it to RH format) as soon as it can, and then attempt to bring up all network interfaces.


This file is laid down by nova, and contains the ssk keys that were provided to nova on instance creation (nova-boot –key ....)


meta.js is populated on the config-drive in response to the user passing “meta flags” (nova boot –meta key=value ...). It is expected to be json formatted.

Version 2

The following criteria are required to as a config drive:

  1. Must be formatted with vfat or iso9660 filesystem or have a filesystem label of config-2
  2. Must be a un-partitioned block device (/dev/vdb, not /dev/vdb1)
  3. The files that will typically be present in the config drive are:
  - 2012-08-10/ or latest/
    - meta_data.json
    - user_data (not mandatory)
  - content/
    - 0000 (referenced content files)
    - 0001
    - ....
  - latest/
    - meta-data.json (not mandatory)

Keys and values

Cloud-init’s behavior can be modified by keys found in the meta.js (version 1 only) file in the following ways.

  values: local, net, pass
  default: pass

This is what indicates if configdrive is a final data source or not. By default it is ‘pass’, meaning this datasource should not be read. Set it to ‘local’ or ‘net’ to stop cloud-init from continuing on to search for other data sources after network config.

The difference between ‘local’ and ‘net’ is that local will not require networking to be up before user-data actions (or boothooks) are run.

  default: iid-dsconfigdrive

This is utilized as the metadata’s instance-id. It should generally be unique, as it is what is used to determine “is this a new instance”.

  default: None

If present, these keys will be used as the public keys for the instance. This value overrides the content in authorized_keys.

Note: it is likely preferable to provide keys via user-data

  default: None

This provides cloud-init user-data. See examples for what all can be present here.


The OpenNebula (ON) datasource supports the contextualization disk.

See contextualization overview, contextualizing VMs and network configuration in the public documentation for more information.

OpenNebula’s virtual machines are contextualized (parametrized) by CD-ROM image, which contains a shell script with custom variables defined on virtual machine start. There are no fixed contextualization variables, but the datasource accepts many used and recommended across the documentation.

Datasource configuration

Datasource accepts following configuration options.

  values: local, net, disabled
  default: net

Tells if this datasource will be processed in ‘local’ (pre-networking) or ‘net’ (post-networking) stage or even completely ‘disabled’.

  default: nobody

Unprivileged system user used for contextualization script processing.

Contextualization disk

The following criteria are required:

  1. Must be formatted with iso9660 filesystem or have a filesystem label of CONTEXT or CDROM
  2. Must contain file with contextualization variables. File is generated by OpenNebula, it has a KEY=’VALUE’ format and can be easily read by bash

Contextualization variables

There are no fixed contextualization variables in OpenNebula, no standard. Following variables were found on various places and revisions of the OpenNebula documentation. Where multiple similar variables are specified, only first found is taken.


Datasource mode configuration override. Values: local, net, disabled.


Static network configuration.


Instance hostname.


If no hostname has been specified, cloud-init will try to create hostname from instance’s IP address in ‘local’ dsmode. In ‘net’ dsmode, cloud-init tries to resolve one of its IP addresses to get hostname.


One or multiple SSH keys (separated by newlines) can be specified.


cloud-init user data.

Example configuration

This example cloud-init configuration (cloud.cfg) enables OpenNebula datasource only in ‘net’ mode.

disable_ec2_metadata: True
datasource_list: ['OpenNebula']
    dsmode: net
    parseuser: nobody

Example VM’s context section

# see

packages: []

- [vdc,none,swap,sw,0,0]
- echo 'Instance has been configured by cloud-init.' | wall
" ]

Alt cloud

The datasource altcloud will be used to pick up user data on RHEVm and vSphere.


For RHEVm v3.0 the userdata is injected into the VM using floppy injection via the RHEVm dashboard “Custom Properties”.

The format of the Custom Properties entry must be:

floppyinject=user-data.txt:<base64 encoded data>

For example to pass a simple bash script:

% cat simple_script.bash
echo "Hello Joe!" >> /tmp/JJV_Joe_out.txt

% base64 < simple_script.bash

To pass this example script to cloud-init running in a RHEVm v3.0 VM set the “Custom Properties” when creating the RHEMv v3.0 VM to:


NOTE: The prefix with file name must be: floppyinject=user-data.txt:

It is also possible to launch a RHEVm v3.0 VM and pass optional user data to it using the Delta Cloud.

For more information on Delta Cloud see:


For VMWare’s vSphere the userdata is injected into the VM as an ISO via the cdrom. This can be done using the vSphere dashboard by connecting an ISO image to the CD/DVD drive.

To pass this example script to cloud-init running in a vSphere VM set the CD/DVD drive when creating the vSphere VM to point to an ISO on the data store.

Note: The ISO must contain the user data.

For example, to pass the same simple_script.bash to vSphere:

Create the ISO

% mkdir my-iso

NOTE: The file name on the ISO must be: user-data.txt

% cp simple_scirpt.bash my-iso/user-data.txt
% genisoimage -o user-data.iso -r my-iso

Verify the ISO

% sudo mkdir /media/vsphere_iso
% sudo mount -o loop JoeV_CI_02.iso /media/vsphere_iso
% cat /media/vsphere_iso/user-data.txt
% sudo umount /media/vsphere_iso

Then, launch the vSphere VM the ISO user-data.iso attached as a CDROM.

It is also possible to launch a vSphere VM and pass optional user data to it using the Delta Cloud.

For more information on Delta Cloud see:

No cloud

The data source NoCloud and NoCloudNet allow the user to provide user-data and meta-data to the instance without running a network service (or even without having a network at all).

You can provide meta-data and user-data to a local vm boot via files on a vfat or iso9660 filesystem. The filesystem volume label must be cidata.

These user-data and meta-data files are expected to be in the following format.


Basically, user-data is simply user-data and meta-data is a yaml formatted file representing what you’d find in the EC2 metadata service.

Given a disk ubuntu 12.04 cloud image in ‘disk.img’, you can create a sufficient disk by following the example below.

## create user-data and meta-data files that will be used
## to modify image on first boot
$ { echo instance-id: iid-local01; echo local-hostname: cloudimg; } > meta-data

$ printf "#cloud-config\npassword: passw0rd\nchpasswd: { expire: False }\nssh_pwauth: True\n" > user-data

## create a disk to attach with some user-data and meta-data
$ genisoimage  -output seed.iso -volid cidata -joliet -rock user-data meta-data

## alternatively, create a vfat filesystem with same files
## $ truncate --size 2M seed.img
## $ mkfs.vfat -n cidata seed.img
## $ mcopy -oi seed.img user-data meta-data ::

## create a new qcow image to boot, backed by your original image
$ qemu-img create -f qcow2 -b disk.img boot-disk.img

## boot the image and login as 'ubuntu' with password 'passw0rd'
## note, passw0rd was set as password through the user-data above,
## there is no password set on these images.
$ kvm -m 256 \
   -net nic -net user,hostfwd=tcp::2222-:22 \
   -drive file=boot-disk.img,if=virtio \
   -drive file=seed.iso,if=virtio

Note: that the instance-id provided (iid-local01 above) is what is used to determine if this is “first boot”. So if you are making updates to user-data you will also have to change that, or start the disk fresh.

Also, you can inject an /etc/network/interfaces file by providing the content for that file in the network-interfaces field of metadata.

Example metadata:

instance-id: iid-abcdefg
network-interfaces: |
  iface eth0 inet static
hostname: myhost



For now see:


Apache CloudStack expose user-data, meta-data, user password and account sshkey thru the Virtual-Router. For more details on meta-data and user-data, refer the CloudStack Administrator Guide.

URLs to access user-data and meta-data from the Virtual Machine. Here is the Virtual Router IP:{metadata type}


Apache CloudStack datasource can be configured as follows:

  CloudStack: {}
  None: {}
  - CloudStack



Vendor Data

The OpenStack metadata server can be configured to serve up vendor data which is available to all instances for consumption. OpenStack vendor data is, generally, a JSON object.

cloud-init will look for configuration in the cloud-init attribute of the vendor data JSON object. cloud-init processes this configuration using the same handlers as user data, so any formats that work for user data should work for vendor data.

For example, configuring the following as vendor data in OpenStack would upgrade packages and install htop on all instances:

{"cloud-init": "#cloud-config\npackage_upgrade: True\npackages:\n - htop"}

For more general information about how cloud-init handles vendor data, including how it can be disabled by users on instances, see


This is the fallback datasource when no other datasource can be selected. It is the equivalent of a empty datasource in that it provides a empty string as userdata and a empty dictionary as metadata. It is useful for testing as well as for when you do not have a need to have an actual datasource to meet your instance requirements (ie you just want to run modules that are not concerned with any external data). It is typically put at the end of the datasource search list so that if all other datasources are not matched, then this one will be so that the user is not left with an inaccessible instance.

Note: the instance id that this datasource provides is iid-datasource-none.