https://wiki.openstack.org/wiki/Heat/ApplicationDeployment
Walk-through in a stack's lifetime
- The user calls the Heat API to create a stack ("heat stack-create ... myStack")
- Heat engine generates a mime multipart data blob that will be consumed by cloud-init
- Heat engine asks nova to create an instance with the cloud-init data
- Nova selects a compute node to run the instance with the cloud-init data
- When the instance boots up, it runs the cloud-init script:
-
- Download the data from the metadata server
- Splits the multiple parts into the /var/lib/cloud/ directory
- Run the different cloud-init parts (resize the root filesystem, set the hostname, install the user's SSH key, etc.)
- Runs the user's script (located at /var/lib/cloud/data/cfn-userdata), it can be any kind of script (Bash, Python, etc.) but at some point this script should call cfn-init.
-
- cfn-init loads /var/lib/cloud/data/cfn-init-data (a copy of the Metadata->AWS::CloudFormation::Init->Config attribute from the AWS template) and can install packages, setup users & groups, create files, etc.
Note that heat-cfntools leverage the boto library and that the boto configuration is stored at /var/lib/heat-cfntools/cfn-boto-cfg on the instance.
Troubleshooting
If you can logged into the instance, look at the following files:
- /var/log/cloud-init.log => cloud-init logs
- /var/log/part-handler.log => logs of the Heat-specific script managing data with content-type=text/x-cfninitdata
- /var/log/heat-provision.log => logs of the user's script including cfn-init logs
You can also query the metadata server for the user's data generated by Heat: