This ansible role does download and install cloudflared
on the host and optionally installs the argo-tunnel as a service.
Breaking changes with 3.0.0
This is a breaking change to reflect the new beahviour of named tunnels
The role should take care of cleanup if you used the role before v.3.0.0. However you have to update the configuration (variables) in your ansible project. I renamed the variables - usually prefixed with
cf_
to make them unique to the role. If they are not unique it may happen that variables using the same name in different roles can have undesired side-effects.
According to 1, in order to create and manage Tunnels, you'll first need to:
- Download and install cloudflared on your machine
- Authenticate cloudflared
Once cloudflared has been installed and authenticated, the process to get your first Tunnel up and running includes 3 high-level steps:
Steps 4-5 are executed once per Tunnel, normally by an administrator, and Step 6 is executed whenever the Tunnel is to be started, normally by the owner of the Tunnel (whom may be different from the administrator).
The role has actually two purposes
The role only takes care of setting up the service on the nodes, i.e. steps 1, 2, 4 and 5 from above, cause
Creating tunnels and enable routing is a task which should be done by an administrator and not the role 1
You can configure one to multiple named tunnels as well as [single service] - even so, with named tunnels you usually only need one daemon. The role actually performs these steps:
-
Download and install binary according to downloads
-
Install/configure the daemon - see Authenticate the daemon
-
For named tunnels a credentials file is created under
{{ cf_credentials_dir }}/{{ tunnel_id }}.json
similar to this{"AccountTag":"{{ account_tag }}","TunnelSecret":"{{ tunnel_secret }}","TunnelID":"{{ tunnel_id }}","TunnelName":"{{ cf_tunnels.key }}"}
-
For each key in
cf_tunnels
create a tunnel config in/etc/cloudflare
The file is named
{{ tunnel }}.yml
and will contain the minimal configuration is as followsnamed tunnels
tunnel: {{ cf_tunnels.key }} credentials-file: {{ cf_credentials_dir }}/{{ tunnel_id }}.json ingress: {{ item.value.ingress }}
single service
hostname: {{ hostname }} url: {{ url }}
Additional parameters are configured using Tunnel configuration params
-
Depending on your init system - controlled by
cf_init_system
- the role does the following-
Systemd
Create a systemd-unit-template
cloudflared@{{ tunnel }}.service
and start an instance for each service in the list ofcf_tunnels
cloudflared tunnel --config {{ tunnel }}.yml
-
Init-V Systems
- Install cloudflared service to
/etc/init.d/{{ systemd_filename }}-{{ tunnel_name }}
- Link Stop-Script to
/etc/init.d/{{ systemd_filename }}-{{ tunnel_name }}
- Link Start-Script to
/etc/init.d/{{ systemd_filename }}-{{ tunnel_name }}
- Install cloudflared service to
-
-
If you use named tunnels the role would also create a dns route.
From where you access your nodes via ssh which is proxied by cloudflared, you need to follow ssh-guide-client. You have to add the following
Host xxx.mycompany.com
ProxyCommand /usr/bin/cloudflared access ssh --hostname %h
You can achieve this configuration if you enable cf_ssh_client_config
. In addition you also need to specify cf_ssh_client_config_group
. So let's assume your inventory looks as follows:
all:
children:
servers:
hosts:
host001:
host002:
If you specify cf_ssh_client_config_group: servers
you would get an entry for host001
and host002
.
none
The following parameters control the installation and/or un-installation
Parameter | Description | Default Value |
---|---|---|
cf_download_baseurl |
Base url for cloudflare binaries |
https://github.com/cloudflare/cloudflared/releases/latest/download/ |
cf_install_only |
Set to true if you only want to install the binary without any configuration or login |
false |
cf_ssh_client_config |
Set to true if you want to configure the proxy configuration for your ssh-guide-client, see SSH Client config |
false |
cf_ssh_client_config_group |
Name of the inventory group for which the ssh proxy config shall be created, see SSH Client config | `` |
cf_force_install |
Set to true if you want to re-install cloudflared . By default the assumption is that cloudflared is running as a service and automatically auto-updates. |
false |
cf_remove_unused_tunnel |
Removes unused cf_tunnels, means cf_tunnels running but not listed in cf_tunnels . |
false |
cf_remove_setup_certificate |
Remove cert.pem after installing the service | false |
cf_credential_file_base |
Folder where to place credential files | /root/.cloudflared/ |
cf_config_dir |
Folder where to place cloudflare configuration files | /etc/cloudflared |
cf_os_package_enable |
Use OS packaging system and Cloudflare package repository (currently just Debian/Ubuntu) | false |
cf_repository_key_url |
If cf_os_package_enable is true, url of the GPG key for the apt repository | https://pkg.cloudflare.com/pubkey.gpg |
cf_repository_key_install_path |
If cf_os_package_enable is true, path where to instal the GPG key for the apt repository | /usr/share/keyrings/cloudflare-main.gpg |
cf_repository |
If cf_os_package_enable is true, url for the Cloudflare apt repository | deb [signed-by={{ cf_repository_key_install_path }}] https://pkg.cloudflare.com/cloudflared {{ ansible_distribution_release }} main |
cf_binary_name |
Name of the cloudflare daemon binary - change only if you know what you are doing | cloudflared |
These are parameters required to create the system service
Parameter | Description | Default Value |
---|---|---|
cf_init_system |
Define which init service to use. Possible values are systemd and initv |
systemd |
cf_systemd_user |
User for systemd service in case cf_init_system: systemd |
root |
cf_systemd_group |
Group for systemd service in case cf_init_system: systemd |
root |
cf_cert_location |
Location of the certificate to be copied - see Authenticate the daemon | - |
cf_cert_content |
Content of the certificate to be copied - see Authenticate the daemon | - |
cf_tunnels |
[Mandatory] List of tunnel-services, each one defining Cloudflare parameters | - |
cf_sysctl_buffer_size_increase |
Increase UDP receive buffer size allowed by the OS (non BSD) - more details | false |
It's recommended to use named tunnels for cf_tunnels
which require Cloudflare named tunnel parameters but you can also use Cloudflare legacy tunnel parameters
...
cf_tunnels:
test:
routes:
dns:
- "{{ inventory_hostname }}"
cidr:
- "192.168.42.0/24"
lb:
- hostname: website.mycompany.com
poolname: bzh-west1.website.mycompany.com
account_tag: !vault....
tunnel_secret: !vault....
tunnel_id: !vault....
ingress:
- hostname: website.mycompany.com
service: http://localhost:1313
- hostname: hello.mycompany.com
service: hello_world
- hostname: ssh.mycompany.com
service: ssh://localhost:22
- service: http_status:404
The key
of the tunnel shall match the of tunnel_id
.
Parameter | Description | Default Value |
---|---|---|
account_tag |
[Mandatory] Account tag from the credentials file generated when creating a tunnel | - |
tunnel_secret |
[Mandatory] Tunnel secret from the credentials file generated when creating a tunnel | - |
tunnel_id |
[Mandatory] Tunnel id from the credentials file generated when creating a tunnel | - |
ingress |
[Mandatory] ingress rules for the tunnel | - |
routes |
List of routes which shall be created. It allows a list for dns -routes at the moment (see example above) |
- |
dns
routes expect a list of CNAME
's to be created as described here. If the CNAME
already exists the task will be skipped but no error thrown. Also only add CNAME
not a FQDN as the FQDN
is determined by cloudlfared
.
private network
routes expect a list of CIDR
's to be created as described here. The playbook loop on the list to execute cloudflared tunnel route ip add {{ cf_cidr_entry }} {{ cf_tunnel.key }}
. If the CIDR
already exists, an error will thrown but ignored.
lb
routes expect a list of existing cloudflared load balancer (plus its pool) to route tunnel on as described here. The playbook loop on the list to execute cloudflared tunnel route lb {{ cf_tunnel.key }} {{ cf_lb_entry.host_name }} {{ cf_lb_entry.pool_name }}
. If the tunnel is already bind into the pool, an ignored error will throw.
As with previous versions of this roles you can use the [single service configuration style][single service]
If you need to proxy traffic to only one local service, you can do so using the config file. As an alternative, you can set up single-service configuration
cf_tunnels:
ssh:
hostname: xxx
url: ssh.mycompany.com
Parameter | Description | Default Value |
---|---|---|
hostname |
[Mandatory]Name or unique | - |
url |
[Mandatory] url to which to connect to config e.g. ssh://localhost:22 or https://localhost:443 |
- |
These are used to configure the parameters per cloudflared service. You still can configure Per-rule configuration for named tunnels as part of the ingress
under cf_tunnels
.
Parameter | Description | Default Value |
---|---|---|
autoupdate_freq |
Autoupdate frequency - see docu | 24h |
no_autoupdate |
Disable periodic check for updates, restarting the server with the new version - see docu | false |
no_tls_verify |
Disables TLS verification of the certificate presented by your origin. Will allow any certificate from the origin to be accepted - see docu | - |
origin_ca_pool |
Path to the CA for the certificate of your origin. This option should be used only if your certificate is not signed by Cloudflare - see docu | - |
origin_server_name |
Hostname that cloudflared should expect from your origin server certificate - see docu |
- |
metrics |
Address to query for usage metrics - see docu | localhost: |
metrics_update_freq |
Frequency to update tunnel metrics - see docu | 5s |
tag |
Custom tags used to identify this tunnel, in format KEY=VALUE - see docu |
- |
protocol |
Specifies the protocol to use for the tunnel - see docu | auto |
loglevel |
Specifies the verbosity of logging. The default "info" is not noisy, but you may wish to run with "warn" in production - see docu | info |
transport_loglevel |
Specifies the verbosity of logs for the transport between cloudflared and the Cloudflare edge. Available levels are: trace , debug , info , warn , error , fatal , panic . Any value below warn produces substantial output and should only be used to debug low-level performance issues and protocol quirks - see docu |
info |
retries |
Maximum number of retries for connection/protocol errors. Retries use exponential backoff (retrying at 1, 2, 4, 8, 16 seconds by default) so increasing this value significantly is not recommended - see docu | 5 |
no_chunked_encoding |
Disables chunked transfer encoding; useful if you are running a WSGI server - see docu | false |
logfile |
Enables writing a logfile for cloudflared - it will still log to the journal | true |
warp_routing |
Allow users to connect to internal services using WARP, details see warp-routing | false |
none
The following example installs an single service for an ssh-tunnel for each server
- hosts: servers
vars:
cf_systemd_user: root
cf_systemd_group: root
cf_cert_location: /home/papanito/cert.pem
services:
ssh:
hostname: "{{ inventory_hostname }}.mycompany.com"
url: ssh://localhost:22
roles:
- papanito.cloudflared
The following example installs an [named tunnel] servers
with an ingress to {{ inventory_hostname }}.mycompany.com
for ssh a hello world if you access hello-{{ inventory_hostname }}.mycompany.com
via the browser
- hosts: servers
remote_user: ansible
become: true
vars:
cf_cert_location: /home/papanito/.cloudflared/cert.mycompany.com.pem
cf_tunnels:
test:
account_tag: !vault...
tunnel_secret: !vault...
tunnel_id: !vault...
routes:
dns:
- "{{ inventory_hostname }}"
- "hello-{{ inventory_hostname }}"
ingress:
- hostname: "hello-{{ inventory_hostname }}.mycompany.com"
service: hello_world
- hostname: "{{ inventory_hostname }}.mycompany.com"
service: ssh://localhost:22
- service: http_status:404
roles:
- papanito.cloudflared
The following example simply downloads cloudflared
on your local machine and configures the ssh-config file:
- hosts: localhost
remote_user: papanito #your local user who has admin
vars:
cf_install_only: true
cf_ssh_client_config: true
cf_ssh_client_config_group: servers
roles:
- papanito.cloudflared
ansible-playbook tests/test.yml -i tests/inventory
According to authenticate-the-cloudflare-daemon when authenticate the daemon, there is a browser window opened or - if this is not possible - then the link has to be put manually. During this time the daemon waits. I could not come up with a solution how to automate this behavior so I came up with the following implementation.
-
if nothing is specified, then ansible calls the
cloudflared login
and will continue when the authentication is done - this makes sens if you use the role to install the daemon locally on your machine and where you have a browser window -
if
cf_cert_location
the certificate is actually copied from thecf_cert_location
, or ifcf_cert_content
is defined then the certificate is created directly from the value stored in it. So you could login once to cloudflare from your master node (where you run ansible) or from a remote location.You can encrypt the
cert.pem
with ansible vault and store it somewhere save.
References:
- downloads - cloudflared download instructions
- ssh-guide - ssh connections with cloudflared
- cli-args - command-line arguments
- config - The configuration file format uses YAML syntax
This is Free Software, released under the terms of the Apache v2 license.
Written by Papanito - Gitlab / Github
[single service]: https://developers.cloudflare.com/cloudflare-one/connections/connect-apps/configuration/ingress#single-service-configuration)