Swift-ansible is a simple way to deploy swift. I personal use it to manage a test cluster, and so isn't full featured, I add features as I require them. for example, there is no EC policy support yet.

The idea, of the deployment is your entire swift cluster is described in an easy to read and use yaml file::

  swift_setup.yml

Which means, this file is an up to date representation of your cluster. The scripts/ directory is non-ansible and contains important scripts to use with the deployment::

  scripts/
    swift_inventory.py
    swift_rings.py

The swift_inventory.py script generates an ansible inventory hosts file, this will one day will just be used as a dynamic inventory, but I haven't gotten around to that yet. Swift_rings.py, creates/updates builder files and creates rings and takes in the swift_setup.yml file. To see how this works see the playbooks/build_swift_rings.yml playbook.

Extra configuration of the swift cluster that isn't managed by the yml file, pipelines for example, and simply managed by editing the swift config files located into the specific swift roles. The swift roles are::

  roles/
    swift_common
    swift_account
    swift_container
    swift_object
    swift_proxy

swift_common, sets up everything common, swift dependencies, checkout the source, installs the source, sets up syslog, swift users, etc. swift_{proxy, account, container, object}, sets up the specific daemons used by swift. So you need to modify the proxy pipeline, you edit::

  roles/swift_proxy/templates/proxy-server.conf.j2

If testing code and you've patched something you can simply deploy the code throughout the cluster with::

   ansible-playbook -i /etc/ansible/hosts /etc/ansible/playbooks/setup_swift.yml -t deploy_code
   ansible-playbook -i /etc/ansible/hosts /etc/ansible/playbooks/restart_swift.yml

Need to run a command and a set of servers, for example I have backed up the swift storage nodes at a certain point for testing, which means I can restore this point by writing a bash script that simply runs::

  ansible container -i /etc/ansible/hosts -m shell -a "swift-init stop all"
  ansible container -i /etc/ansible/hosts -m shell -a "rm -Rf /mnt/sda/*"
  ansible container -i /etc/ansible/hosts -m shell -a "rm -Rf /mnt/sdb/*"
  ansible container -i /etc/ansible/hosts -m shell -a "rm -Rf /mnt/sdc/*"
  ansible container -i /etc/ansible/hosts -m shell -a "rm -Rf /mnt/sdd/*"
  ansible container -i /etc/ansible/hosts -m shell -a "tar -xvf /root/cont_300k.tar -C /mnt/"
  ansible-playbook -i /etc/ansible/hosts /etc/ansible/playbooks/restart_swift.yml

NOTE: in my test cluster my storage nodes are members of the container, account and object groups, so by using the container group (ansible container) above will hit all my storage nodes.

To build and deploy the rings:

   mkdir rings && cd rings
   python ../scripts/swift_rings.py -s ../swift_setup.yml
   cd -
   ansible-playbook -i hosts deploy_swift_rings.yml

Swift-Ansible can also be installed on VMs created by Vagrant, currently this will require the following manual steps::

• vagrant up <= This will create 6 VMs by default (see Vagrantfile)
• cp swift_setup.yml.example swift_setup.yml
• python scripts/swift_inventory.py -s swift_setup.yml
• ssh-copy-id [email protected] <= repeat for each newly created VM
• ansible-playbook -i hosts configure_swift.yml
• mkdir rings && cd rings
• python ../scripts/swift_rings.py -s ../swift_setup.yml

= Statsd role = There is now a statsd role, which will deploy collectd, influxdb and grafana. It sets up collectd to listen for statdsd and write out to influxdb.

By default it will use the default ports, but check roles/statsd/vars/main.yml if you want to change anything.

Once complete, you will need to add the influxdb datasource in grafana:

• browse to: http://<statsd_host>:3000
• Default login in admin/admin (you can change this with the grafana-cli admin reset-admin-password <new-password> command.
• Create an influxdb datasource where:
  • hosts: http://localhost:8086
  • database: collectd

NOTE: You can see if datapoints are making it into influxdb with:

curl -G http://127.0.0.1:8086/query?pretty=true --data-urlencode "db=collectd" --data-urlencode "q=show series"