This is an Ansible role for use with Tor - https://www.torproject.org/

I hope that relay operators will find this useful for deploying and maintaining large numbers of Tor relays and bridges with finesse, concurrency and idempotency!

This ansible role can help you reduce the complexity to a single command for deploying one or more tor relays or tor bridges with or without obfsproxy.

Here I assume the user has setup their ansible project directory according to the best practices directory-layout specified here:

http://docs.ansible.com/playbooks_best_practices.html#directory-layout

...and simply run a command like this: ansible-playbook -i production tor-relays.yml

In this case I'm running the tor-relays.yml playbook against the "production" inventory file. This tor-relays playbook might specify a host group called tor-relays... which is defined in the inventory file. I could have many other host groups defined in the inventory as well such as: tor-exit-relays, tor-bridges, tor-bananaphone-bridges, tor-hidden-tahoe-storage-nodes etc.

Works on Debian and Ubuntu. I've tried it... so I know. =-)

tor_distribution_release should be set to the desired distribution of the Tor Project's APT repo - http://deb.torproject.org/torproject.org

Perhaps many debian users will want to specify "wheezy"... however if you intend to operate a tor bridge then you unless you probably want "tor-experimental-0.2.5.x-wheezy" so that you can have ServerTransportOptions in your torrc.

tor_obfsproxy_home variable should set when you want to use obfsproxy with your bridge configuration. Perhaps I should change this to be a boolean variable names tor_run_obfsproxy... and then set a reasonable default for the obfsproxy python virtual env directory?

tor_wait_for_hidden_services can be set to yes if you would like the ansible-tor role to wait for the newly created tor hidden services to start. It does so by waiting for the tor hidden service hostname file to appear.

---
- hosts: tor-relays
  user: human
  connection: ssh
  roles:
# XXX uber-paranoid openssh ansible role?
#    - { role: ansible-openssh-hardened,
#        backports_url: "http://ftp.de.debian.org/debian/",
#        backports_distribution_release: "wheezy-backports",
#        ssh_admin_ed25519pubkey_path: "/home/amnesia/.ssh/id_ed25519.pub",
#        sudo: yes
#      }
#    - { role: ansible-tlsdate,
#        remove_ntp: yes,
#        sudo: yes
#      }
    - { role: ansible-tor,
        tor_distribution_release: "tor-experimental-0.2.5.x-wheezy",
        tor_BridgeRelay: 1,
        tor_PublishServerDescriptor: "bridge",
        tor_ExtORPort: "auto",
        tor_ORPort: 9001,
        tor_ServerTransportPlugin: "obfs4 exec /usr/bin/obfs4proxy",
        tor_ExitPolicy: "reject *:*",
        tor_obfs4proxy_enabled: True,
        sudo: yes
      }

Note that the ansible-tlsdate role is also not strictly necessary... however Tor does need accurate time and I think it is much better to use tldated instead of ntpd. See here: https://github.com/david415/ansible-tlsdate

Furthermore it is also a good idea to have a hardened openssh-server configuration; that supports the new ed25515 key exchange + the new polychacha1305 DJB-inspired crypto transport. See here: https://github.com/david415/ansible-openssh-hardened

This playbook installs and fully configures a scramblesuit ( http://www.cs.kau.se/philwint/scramblesuit/ ) tor bridge using the latest obfsproxy available to pip (installs into a python virtualenv).

---

- hosts: tor-bridges
  roles:
    - { role: ansible-role-firewall,
        firewall_allowed_tcp_ports: [ 22, 4703 ],
        sudo: yes
      }
    - { role: david415.ansible-tor,
        tor_distribution_release: "wheezy",
        tor_BridgeRelay: 1,
        tor_PublishServerDescriptor: "bridge",
        tor_obfsproxy_home: "/home/ansible",
        tor_ORPort: 9001,
        tor_ServerTransportPlugin: "scramblesuit exec {{ tor_obfsproxy_home }}/{{ tor_obfsproxy_virtenv }}/bin/obfsproxy --log-min-severity=info --log-file=/var/log/tor/obfsproxy.log managed",
        tor_ServerTransportListenAddr: "scramblesuit 0.0.0.0:4703",
        tor_ExitPolicy: "reject *:*",
        sudo: yes
      }

You should also feel free to apply iptables rulesets... however this also is not strictly necessary but at times might be a good idea.

This playbook demonstrates configuring a tor bridge with an obfsproxy installed from my git repo so that the bananaphone pluggable transport is available (it has not been merged upstream).

Bananaphone provides tor over markov chains! If you have sensitive or interesting documents then please consider operating a bananaphone bridge utilizing these text corpuses. Read about the bananaphone pluggable transport for tor - http://bananaphone.io/

---

- hosts: tor-bridges
  roles:
    - { role: david415.ansible-tor,
        tor_distribution_release: "tor-experimental-0.2.5.x-wheezy",
        tor_BridgeRelay: 1,
        tor_PublishServerDescriptor: "bridge",
        tor_obfsproxy_home: "/home/ansible",
        tor_ORPort: 9001,
        tor_obfsproxy_git_url: "git+https://github.com/david415/obfsproxy.git",
        tor_ServerTransportPlugin: "bananaphone exec {{ tor_obfsproxy_home }}/{{ tor_obfsproxy_virtenv }}/bin/obfsproxy --log-min-severity=info --log-file=/var/log/tor/obfsproxy.log managed",
        tor_ServerTransportOptions: "bananaphone corpus=/usr/share/dict/words encodingSpec=words,sha1,4 modelName=markov order=1",
        tor_ServerTransportListenAddr: "bananaphone 0.0.0.0:4703",
        tor_ExitPolicy: "reject *:*",
        sudo: yes
      }

This example playbook sets up tor relays with hidden service for ssh...

This playbook demonstrates waiting for the hidden services to be created... by awaiting the existence of the tor hidden service hostname files. This happens when the role variable "tor_wait_for_hidden_services" is set to yes.

This feature could be useful when configuring other services that depend on knowing the hidden service's onion address... such as my ansible-tahoe-lafs role: https://github.com/david415/ansible-tahoe-lafs

Read about Tahoe-LAFS here: https://tahoe-lafs.org/trac/tahoe-lafs Read about tor hidden services here: https://www.torproject.org/docs/tor-hidden-service.html.en

---
- hosts: tor-relays
  user: ansible
  connection: ssh
  vars:
    relay_hidden_services_parent_dir: "/var/lib/tor/services"
    relay_hidden_services: [ { dir: "hidden_ssh",
                               ports: [ { virtport: "22",
                                 target: "localhost:22" } ] }
    ]
  roles:
    - { role: ansible-role-firewall,
        firewall_allowed_tcp_ports: [ 22, 9001 ],
        sudo: yes
      }
    - { role: david415.ansible-tor,
        tor_distribution_release: "wheezy",
        tor_ExitPolicy: "reject *:*",
        tor_hidden_services: "{{ relay_hidden_services }}",
        tor_hidden_services_parent_dir: "{{ relay_hidden_services_parent_dir }}",
        tor_wait_for_hidden_services: yes,
        sudo: yes
      }

polytorus-ansibilus.yml:

---
- hosts: tor-relays
  roles:
    - { role: david415.ansible-tor,
        tor_distribution_release: "wheezy",
        tor_ExitPolicy: "reject *:*",
        sudo: yes
      }

A simple playbook like this can be used to deploy many instances of tor on many servers. You can configure multiple tor instances using the host_vars file for each host.

Here's what an example host_vars file looks like (with rfc1918 ip addrs):

host_vars/192.168.1.1:

tor_Nickname: [ "ScratchMaster" ]
proc_instances: [ {
name: "relay1",
tor_ORPort: ["192.168.1.1:9002"],
tor_SOCKSPort: ["8041"]
},
{
name: "relay2",
tor_ORPort: ["192.168.1.2:9002"],
tor_SOCKSPort: ["8042"]
},
{
name: "relay3",
tor_ORPort: ["192.168.1.3:9002"],
tor_SOCKSPort: ["8043"]
}]

In the above example playbook, all the role variables get applied to all tor instances. If you want to control the role variables for a specific host then you must use that host's host_vars file.

Note: when this role is used in "multi-tor process mode"... meaning that if the proc_instances variable is defined... then the torrc template will set reasonable defaults for these torrc options: User, PidFile, Log and DataDirectory.

This next example is NOT very practical because it can only be used with a host inventory with one host! If it were to be used with multiple hosts then their torrc files would contain the same IP addresses.

---
- hosts: tor-relays
  roles:
    - { role: david415.ansible-tor,
        tor_distribution_release: "wheezy",
        tor_ExitPolicy: "reject *:*",
        tor_instance_parent_dir: "/etc/tor/instances",
        proc_instances: [ {
                          name: "relay1",
                          tor_ORPort: ["192.168.1.1:9002"],
                          tor_SocksPort: ["8041"]
                        },
                        {
                          name: "relay2",
                          tor_ORPort: ["192.168.1.2:9002"],
                          tor_SocksPort: ["8042"]
                        },
                        {
                          name: "relay3",
                          tor_ORPort: ["192.168.1.3:9002"],
                          tor_SocksPort: ["8043"]
                        }],
        sudo: yes
      }

torrc may have options set from host_vars/group_vars and also set from role variables.

The host_vars can set arbitrary torrc configuration options however the role variables currently support a small subset of the torrc options at the moment... it's a work in progress; refer to the templates/torrc for a more detailed overview.

The host_vars variable names must begin with "tor_"; Here's an example setting "Nickname":

tor_Nickname: [ "OnionRobot" ]

The dictionary value is a list because in some cases you may want to specify multiple lines in the torrc that begin with the dictionary key.

MIT

https://github.com/david415/ansible-tor/issues