Ansible requires a "command machine" or "local machine" or "orchestrator machine" to run on. This can be your laptop or any machine that can run ansible. (It does not have to be part of the cloud network that hosts your servers.)

Use the official Ansible installation guide to install Ansible. Here are a few examples on basic installation commands:

Ubuntu/Debian:

sudo apt-get install ansible

CentOS/RedHat:

sudo yum install epel-release
sudo yum install ansible

Mac OSX: If you have (Homebrew)[https://brew.sh] installed, then it's simply

brew install ansible

If not, you can install it using pip:

sudo easy_install pip
sudo pip install ansible

To make life easier, you can start an SSH Agent and load your SSH key(s). This way ansible will have an uninterrupted way of connecting to your servers.

ssh-agent > ~/.ssh/ssh.env
source ~/.ssh/ssh.env

ssh-add private.key

Subsequently, as long as the agent is running, you can use source ~/.ssh/ssh.env to load the keys to the current session. Note: On Mac OSX, you can add the -K option to ssh-add to store the passphrase in your keychain. The security of this feature is debated but it is convenient.

Optional cloud dependencies

If you are using a cloud provider to host your servers, you need the below dependencies installed on your local machine.

DigitalOcean inventory dependencies:

Ubuntu/Debian:

sudo apt-get install python-pip
sudo pip install dopy

CentOS/RedHat:

sudo yum install python-pip
sudo pip install dopy

Mac OSX:

sudo pip install dopy

Amazon AWS inventory dependencies:

Ubuntu/Debian:

sudo apt-get install python-boto

CentOS/RedHat:

sudo yum install python-boto

Mac OSX:

sudo pip install boto

Refreshing the DigitalOcean inventory

If you just finished creating droplets, the local DigitalOcean inventory cache is not up-to-date. To refresh it, run:

DO_API_TOKEN="<The API token received from DigitalOcean>"
python -u inventory/digital_ocean.py --refresh-cache 1> /dev/null

Refreshing the Amazon AWS inventory

If you just finished creating Amazon AWS EC2 instances, the local AWS inventory cache is not up-to-date. To refresh it, run:

AWS_ACCESS_KEY_ID='<The API access key ID received from Amazon>'
AWS_SECRET_ACCESS_KEY='<The API secret access key received from Amazon>'
python -u inventory/ec2.py --refresh-cache 1> /dev/null

Note: you don't need the access key and secret key set, if you are running ansible on an Amazon AMI instance with the proper IAM permissions set.

Running the playbooks

The playbooks are locked down to only run if the environment variable TF_VAR_TESTNET_NAME is populated. This is a precaution so you don't accidentally run the playbook on all your servers.

The variable TF_VAR_TESTNET_NAME contains the testnet name which ansible translates into an ansible group. If you used Terraform to create the servers, it was the testnet name used there.

If the playbook cannot connect to the servers because of public key denial, your SSH Agent is not set up properly. Alternatively you can add the SSH key to ansible using the --private-key option.

If you need to connect to the nodes as root but your local username is different, use the ansible option -u root to tell ansible to connect to the servers and authenticate as the root user.

If you secured your server and you need to sudo for root access, use the the -b or --become option to tell ansible to sudo to root after connecting to the server. In the Terraform-DigitalOcean example, if you created the ec2-user by adding the noroot=true option (or if you are simply on Amazon AWS), you need to add the options -u ec2-user -b to ansible to tell it to connect as the ec2-user and then sudo to root to run the playbook.

DigitalOcean

DO_API_TOKEN="<The API token received from DigitalOcean>"
TF_VAR_TESTNET_NAME="testnet-servers"
ansible-playbook -i inventory/digital_ocean.py install.yml -e service=basecoin

Amazon AWS

AWS_ACCESS_KEY_ID='<The API access key ID received from Amazon>'
AWS_SECRET_ACCESS_KEY='<The API secret access key received from Amazon>'
TF_VAR_TESTNET_NAME="testnet-servers"
ansible-playbook -i inventory/ec2.py install.yml -e service=basecoin

Installing custom versions

By default ansible installs the tendermint, basecoin or ethermint binary versions from the latest release in the repository. If you build your own version of the binaries, you can tell ansible to install that instead.

GOPATH="<your go path>"
go get -u github.com/tendermint/basecoin/cmd/basecoin

DO_API_TOKEN="<The API token received from DigitalOcean>"
TF_VAR_TESTNET_NAME="testnet-servers"
ansible-playbook -i inventory/digital_ocean.py install.yml -e service=basecoin -e release_install=false

Alternatively you can change the variable settings in group_vars/all.

Other commands and roles

There are few extra playbooks to make life easier managing your servers.

install.yml - Install basecoin or ethermint applications. (Tendermint gets installed automatically.) Use the service parameter to define which application to install. Defaults to basecoin.
reset.yml - Stop the application, reset the configuration and data, then start the application again. You need to pass -e service=<servicename>, like -e service=basecoin. It will restart the underlying tendermint application too.
restart.yml - Restart a service on all nodes. You need to pass -e service=<servicename>, like -e service=basecoin. It will restart the underlying tendermint application too.
stop.yml - Stop the application. You need to pass -e service=<servicename>.
status.yml - Check the service status and print it. You need to pass -e service=<servicename>.
start.yml - Start the application. You need to pass -e service=<servicename>.
ubuntu16-patch.yml - Ubuntu 16.04 does not have the minimum required python package installed to be able to run ansible. If you are using ubuntu, run this playbook first on the target machines. This will install the python pacakge that is required for ansible to work correctly on the remote nodes.
upgrade.yml - Upgrade the service on your testnet. It will stop the service and restart it at the end. It will only work if the upgraded version is backward compatible with the installed version.
upgrade-reset.yml - Upgrade the service on your testnet and reset the database. It will stop the service and restart it at the end. It will work for upgrades where the new version is not backward-compatible with the installed version - however it will reset the testnet to its default.

The roles are self-sufficient under the roles/ folder.

install - install the application defined in the service parameter. It can install release packages and update them with custom-compiled binaries.
unsafe_reset - delete the database for a service, including the tendermint database.
config - configure the application defined in service. It also configures the underlying tendermint service. Check group_vars/all for options.
stop - stop an application. Requires the service parameter set.
status - check the status of an application. Requires the service parameter set.
start - start an application. Requires the service parameter set.

Default variables

Default variables are documented under group_vars/all. You can the parameters there to deploy a previously created genesis.json file (instead of dynamically creating it) or if you want to deploy custom built binaries instead of deploying a released version.