Skip to content

Latest commit

 

History

History
205 lines (131 loc) · 9.22 KB

README.md

File metadata and controls

205 lines (131 loc) · 9.22 KB
Raw

ansible-pcd

ansible-pcd provides ansible roles and structural conventions to easily automate (even massive) *nix infrastructure.

roles are organized by system, service, or application and tasks are tagged as either prepare, configure, or deploy. this is scalable and drastically reduces execution time by visiting only intended tasks.

the project has the following goals;

  • remain distribution agnostic (support Debian, RedHat, &c out-of-box)
  • encourage reusable automation components, avoid redundancy
  • embrace ansible-best-practices -- remain understandable and community friendly

See the ginas project as an alternative.

quick start

  • ensure ansible 1.6+
  • clone or download ansible pcd to a directory (now known as PCDROOT)
  • create your private directory (which holds keys & passwords) 
    cp -a /PCDROOT/private.sample /PCDROOT/private
    • you may define an alternative location of the private directory by overriding the PCD_PRIVATE_DIR inventory variable.
    • remove /private from /PCDROOT/.gitignore if you want to check-in private keys & passwords.
      • you may use the vault to encrypt files
  • create your host inventory, use /PCDROOT/inventory/iceburg.hosts as a reference
  • configure your private directory, set passwords, ssl certs, and add ssh keys for connecting to hosts 
    ssh-keygen -t rsa -f /PCDROOT/private/keys/ENV+root.key
    • see /PCDROOT/private.sample/vars files as reference
  • configure hosts using inventory variables
    • see /PCDROOT/inventory/group_vars files and /PCDROOT/inventory/host_vars files as reference
    • PCD roles provide overridable defaults - visit the roles to get an idea of what you can configure.
  • modify /PCDROOT/site.yml to your needs and execute 
    ansible-playbook -i inventory/my.hosts site.yml
    • use the --ask-pass flag to connect to hosts that have not yet been provisioned via the pcd_system role.

usage

The pcd_*.yml playbooks are used to apply a single role to specified host(s). They are executed via ansible-playbook, and you pass paramaters to indicate the desired role, tasks, and target hosts to execute upon.

For example, you can use the pcd_service.yml playbook to quickly provision mysql on all hosts in the dbservers group via ansible-playbook -i inventory/iceburg.hosts pcd_service.yml --limit=dbservers -e PCD_ROLE=mysql

Running the pcd_*.yml playbooks without tags will execute all tasks. As such, the above example will execute prepare, configure, and deploy tasks in that order.

  • prepare tasks are ideally run once per host. They install packages, add users, create directories, &c.
  • configure tasks are run more often. They setup cron jobs, template configuration files, set timezone, &c.
  • deploy tasks are run most often, and limited to applications and sites. Run to deploy code changes.

Many pcd roles provide default configuration variables meant to be overriden. These are typically UPPERCASED. Configure to your liking by redefining them in your inventory variables.

It's best to define your own infrastructure and the applications/services that run on it. You do this through standard ansible playbooks. Included is an example that provisions an entire environemnt and deploys websites. See:

Other examples provided;

examples


# provision all hosts in your inventory file
ansible-playbook -i inventory/iceburg.hosts pcd_system.yml 


# reconfigure the apache service on a specific host
ansible-playbook -i inventory/iceburg.hosts pcd_service.yml -t configure -l ocean-1.iceburg.net -e PCD_ROLE=apache"


# apply a site to all webservers (you will be prompted for site name and org if not passed)
ansible-playbook -i inventory/iceburg.hosts pcd_site.yml -l webservers


# deploy (in this case; git checkout) www.iceburg.net to ocean-1.iceburg.net
ansible-playbook -i inventory/iceburg.hosts pcd_site.yml -t deploy -l ocean-1.iceburg.net --extra-vars="PCD_SITE_NAME=www.iceburg.net PCD_SITE_ORG=iceburg"

connecting, initial provisioning

The pcd-systems roles provision hosts with a consistent environment. They ensure the root user's authorized keys for ansible to connect, set the fqdn properly, install a common set of packages, and tighten security.

When connecting to a host for the first time (that doesn't yet have an authorized key for the root user), pass the --ask-pass flag to ansible-playbook.

By default, ansible-pcd connects to hosts as the root user using ssh keys from the /PCDROOT/private/keys directory. The ansible_ssh_private_key_file inventory variable determines the key used, and defaults to {{ PCD_KEYS_DIR }}/{{ PCD_DEFAULT_ORG }}+{{ ansible_ssh_user }}.key.

For instance, if we're connecting to a host belonging to the chicago-east organization, ansible would select <pcd-root>/private/keys/chicago-east+root.key. Again, all this is configured via inventory variables.

status

development

ansible-pcd is under development. there may be breaking api changes in the near future.

current functionality

at this time ansible-pcd provides a "webhost in a box"

hosts are provisioned with a consistent, secure environment. services are configurable per host using playbooks and inventory.

websites are defined in YAML with built-in, real-world conveniences;

  • git based sites (deployment via shallow checkout)
  • wordpress/silverstripe/&c rewrites
  • pcd.mysql integration (create user + database)
  • pcd.awstats integration (registers logfile to be analyzed)
  • pcd.backup integration (site assets/uploads > clound storage via s3ql)

Also featured are useful infrastructure roles for monitoring, remote filesystems, varnish caching, and VPN connectivity.

more to come, please contribute!

contributing

ansible-pcd is licensed under the GPLv3 , the same as ansible.

conventions

  • role tasks must be tagged as either prepare, configure, or deploy. to reduce redundancy and add convenience, tasks are typically separated into tagged includes.
# example pcd-role/tasks/main.yml

- { include: prepare.yml, tags: ['prepare'] }
- { include: configure.yml, tags: ['configure'] }
- { include: deploy.yml, tags: ['deploy'], sudo: True, sudo_user: "{{ HTTPD_USER }}" }

# example pcd-role/tasks/prepare.yml

- debug: msg="I am executed when the `prepare` tag is passed."

  • If a variable is a good candidate to be shared by other roles, place it in pcd-common/defaults/main.yml. If it is distribution/OS specific, place appropriate values in pcd-common/vars.

  • Follow the edX project standard and CAPITALIZE the names of variables likely to be overriden/configured by users. Place them at the top of your defaults/main.yml.

  • Please support both Debian and RedHat OS families

support

Report issues to the github issue tracker.

For support, please post to stackoverflow using the ansible-pcd tag:

developer-todo

  • ncurses based UI for applying roles
  • map system uuid => fqdn to more easily identify remote backups
  • make awstats apache/nginx agnostic, set inventory preference for nginx|apache
  • iptables role to compliment pptpd and openvpn [ferm!]
  • backup /etc/ansible / migrate to rclone or rsnapshot based backups?
  • refactor / rethink shared tasks. incompatible w/ ansible 1.8... ACK!