This project provides a framework for migrating applications from an OS (VM or bare-metal) into a container running on a likely newer platform. This is being referred to here as lift and shift. Note that this term is widely used to describe different types of application migrations, but in general refers to migrating applications to a cloud platform rather than refactoring the application to take advantage of the native cloud platform. This is a quicker approach but may require more work to maintain longer term. However, in the case of lifting an application from a legacy OS that may not be taking full advantage of newer hardware, for example, this could be very advantageous.
Lift and shift could also apply to other migration types. For example, lift and shift of the entire OS in a VM and creating a container from this which would more closely resemble a pet container. It could also be possible to lift and shift application-specific code using this approach only, and create a container with the required framework in the case of something using JBoss EAP for example.
Before running the playbook, you must first define a profile for each of the applications to look for in inventory/profiles.yml. This is a list of applications with attributes such as services, packages, files and directories that comprise that application. The example used here is wordpress which uses mysql, httpd services, php-mysql and other packages (see profiles.yml for details).
Next, create matching tasks for this profile in roles/lift/tasks/{profile}.yml and roles/shift/tasks/{profile}.yml roles using the same name defined in profiles.yml. Refer to the wordpress.yml in each of the roles/{lift,shift}/tasks/ directories for examples and see below for a description of each role. Note that the shift role also requires files and templates to define the Dockerfile and any supporting scripts to start the container and restore the application.
Lastly, define a hosts file with source hosts to scan and a destination host to lift and shift the matching application to. This initial prototype assumes an Atomic Host but as it simply uses docker module it could be any system running docker. Future plans include OpenShift support.
This role is called first, unless the perform_introspection is overridden to False. In this case you must explicitly assign a profile type to each source host in the inventory using the variable profile. Now the playbook is ready to be run. First read about the roles before doing so.
This role is called next and is where the matching roles/lift/tasks/{profile}.yml is called. This is where each defined profile attribute should be backed up depending on the type of attribute. For example, a mysqldump for a database. Some attributes may not require an action at this stage yet such as any pacakges, but may be used in the shift role.
This role moves data from the lift role to the destination host, then pulls the appropriate docker image, does a docker build and starts the container. The Dockerfile and any supporting files should be provided as templates or files in the role with the matching profile name. See roles/shift/files/wordpress and roles/shift/templates/wordpress for examples. Also see https://github.com/CentOS/CentOS-Dockerfiles for additional ideas as the wordpress example used here was modified from that repo.
With the profile and supporting tasks defined, simply run the playbook as such:
$ ansible-playbook shifter.yml
To use the included wordpress lift and shift profile, create a simple wordpress instance on a CentOS 6 or RHEL 6 system using this site as a guide: https://www.theurbanpenguin.com/creating-wordpress-site-on-rhel-7-1/
Be sure to substitute mysql for mariadb. The database and wordpress password are set in the roles/shift/templates/wordpress-start.sh and hard coded as shifter but can be easily made dynamic via a variable.
I hope to add more example application profiles, include some middleware examples and target OpenShift as a destination.