Using Enough

Knowledge

Enough is based on Ansible, and Debian GNU/Linux. When running in the cloud it relies on OpenStack. When running on physical hardware, it relies on libvirt. In both cases the Debian GNU/Linux based virtual machines are provisioned with Ansible playbooks.

Each service adds more components such as Let’s encrypt, Docker or Snap. And they also implement concepts such as Virtual Private Network, Reverse Proxy or Certificate Authority.

If all goes well, Enough can be used even if the user knows nothing more than what is in this guide. When something goes wrong, fixing the problem or understanding the cause will require intimate knowledge about all of the above. If that happens, it is best to start a discussion in the forum and ask for help.

Using the cloud or physical machines

Enough can be 100% in the cloud (using OpenStack), 100% on physical machines (using libvirt), or a mixture of both connected together with a VPN.

Installation of the Enough CLI

  • Install Docker.
  • Copy clouds.yml in ~/.enough/example.com/inventory/group_vars/all/clouds.yml and edit to add password: next to auth_url: under the production: name.
  • Add the enough CLI to ~/.bashrc:
eval "$(docker run --rm enoughcommunity/enough:latest install)"
  • Verify that it works:
enough --version

Upgrade

To upgrade to the latest Enough version:

$ docker pull enoughcommunity/enough:latest
$ eval "$(docker run --rm enoughcommunity/enough:latest install)"

OpenStack Enough instance

Requirements

  • The clouds.yml credentials for either OVH or Fuga.
  • The Orchestration API available: Heat, at least heat_template_version: 2016-10-14/newton.
  • A public IPv4 per virtual machine (not a floating IP but a direct IP). IPv6 isn’t supported yet.
  • In order to deploy every available services: 15 virtual machines.
  • The flavors must be provided with an attached root disk by default (not an explicit block storage) most of the virtual machines use 2Go RAM, some hosts/tests require 4Go or 8Go RAM.
  • A Debian GNU/Linux stable image.

Create the DNS name server

Assuming the domain name (example.com for example) is registered, it must be delegated to a dedicated name server before any service can be created by Enough:

enough --domain example.com service create bind

Upon successfull completion, a machine named bind-host exists and its public IP must be used as a GLUE record.

$ enough --domain example.com openstack server list
+---------------+--------------+--------+-----------------------+-----------+--------+
| ID            | Name         | Status | Networks              | Image     | Flavor |
+---------------+--------------+--------+-----------------------+-----------+--------+
| 2b9a1bda-c2c0 | bind-host    | ACTIVE | Ext-Net=51.178.60.121 | Debian 10 | s1-2   |
+---------------+--------------+--------+-----------------------+-----------+--------+

To verify the DNS running at bind-host works as expected:

$ dig @ip-of-the-bind-host bind.example.com
$ enough --domain example.com ssh bind-host

It can then be used to instruct the registrar of example.com to delegate all domain name resolutions to this IP. The exact method for performing this DNS delegation depends on the registrar (Gandi is different from OVH, etc.). But it needs to be done only once.

Note

It will take time for the delegation to be effective. It can be as quick as one hour but delays from 24h to 72h are not uncommon.

To verify the delegation is effective:

getent hosts bind.example.com

libvirt Enough instance

Requirements

  • A physical machine with Debian GNU/Linux stable (the IP of the machine is 192.168.1.19 in the examples below)
  • A debian user with passwordless ssh access from the machine where the Enough CLI is installed:
$ enough --domain lan.example.com info
$ ssh-copy-id -i ~/.enough/enough.lan/infrastructure_key.pub debian@192.168.1.19
  • Allow debian passwordless sudo access:
$ ssh debian@192.168.1.19
$ su
$ echo 'debian ALL=(ALL) NOPASSWD:ALL' > /etc/sudoers.d/debian

Setup libvirt

The libvirt daemon and tools must be installed as follows:

enough --domain lan.example.com libvirt install 192.168.1.19

Note

If a docker was installed on the hypervisor, the daemon must be restarted with systemctl restart docker so that it re-installs its iptables rules because the installation of the libvirt daemon discards them.

Create the DNS name server

enough --domain lan.example.com service create bind --driver libvirt

Connecting libvirt and OpenStack Enough instances

The OpenStack Enough instance runs a VPN server to which a host of the libvirt Enough instance connects as a client. Routes are setup so that all hosts in both the OpenStack and libvirt Enough instances can communicate.

The libvirt hypervisor is also a client of the VPN so that system administration can be done remotely.

Configuring the libvirt hypervisor as a VPN client

Assuming the libvirt hypervisor client is defined in ~/.enough/example.com/inventory/group_vars/all/openvpn.yml as follows:

---
#
#############################################
#
# List of active openvpn clients
#
openvpn_active_clients:
  - hypervisor

After running the VPN service as follows:

$ enough --domain example.com service create --host bind-host openvpn

The file ~/.enough/example.com/openvpn/hypervisor.tar.gz is created and must be manually copied from example.com to lan.example.com in the file ~/.enough/lan.example.com/hypervisor.tar.gz.

It can then be used to configure the lan.example.com hypervisor as a VPN client of example.com as follows:

$ enough --domain lan.example.com libvirt install \
     --vpn hypervisor.tar.gz 192.168.1.19

Finally it must be started as follows:

$ systemctl start openvpn-client@hypervisor

Configuring the libvirt bind-host as a VPN client

Assuming the lan client is defined in ~/.enough/example.com/inventory/group_vars/all/openvpn.yml as follows:

---
#
#############################################
# List of active openvpn clients
#
openvpn_active_clients:
  - lan

After running the VPN service as follows:

$ enough --domain example.com service create --host bind-host openvpn

The file ~/.enough/example.com/openvpn/lan.tar.gz is created and must be manually copied from example.com to lan.example.com in the file ~/.enough/lan.example.com/lan.tar.gz.

The lan client is associated with some-host in ~/.enough/example.com/inventory/host_vars/some-host/openvpn.yml as follows:

---
openvpnclient_name: lan

Finally it can be setup and run on some-host as follows:

$ enough --domain lan.example.com service create --driver libvirt \
    --host some-host openvpnclient

Create or update a service

The following services are available:

As an example, the cloud service can be created as follows, with OpenStack:

enough --domain example.com service create cloud

Note

If the command fails, because of a network failure or any other reason, it is safe to run it again. It is idempotent.

When it completes successfully, it is possible to login https://cloud.example.com with user admin and password mynextcloud.

If running with libvirt instead of OpenStack, the –driver libvirt argument must be explicitly provided as follows:

enough --domain example.com service create --driver libvirt cloud

Restore a service

Stateless services such as bind do not need backup: they can be rebuilt from scratch if the machine hosting them fails. For instance, if bind-host is lost:

$ enough --domain example.com host create bind-host
$ enough --domain example.com playbook

However, most services such as file sharing and translations rely on persistent information that are located in a encrypted volume attached to the machine. A daily backup is made in case a file is inadvertendly lost.

OpenStack infrastructure services and access

Networks

By default all hosts are connected to two networks: one with a public IP and the other with a private IP. This can be changed by setting the network_internal_only variable in ~/.enough/example.com/inventory/group_vars/all/network.yml, using this example.

The default can also be changed for a given host (for instance weblate-host) by setting the desired value in the ~/.enough/example.com/inventory/host_vars/weblate-host/network.yml file.

VPN

A VPN can optionally be installed for clients to access hosts that do not have public IPs.

A host with a public IP must be chosen to deploy the VPN. For instance bind-host by adding the following to ~/.enough/example.com/inventory/services.yml:

openvpn-service-group:
  hosts:
    bind-host:

It can then be created with:

enough --domain example.com service create openvpn

The certificates for clients to connect to the VPN will be created from the list in the openvpn_active_clients variable in ~/.enough/example.com/inventory/group_vars/all/openvpn.yml, using this example.

For each name in the openvpn_active_clients list, a .tar.gz file will be created in the ~/.enough/example.com/openvpn/ directory. For instance, for

---
openvpn_active_clients:
 - loic

The file ~/.enough/example.com/openvpn/loic.tar.gz will be created and contains OpenVPN credentials. The specific instructions to use them depends on the client.

Certificates

By default certificates are obtained from Let’s Encrypt when using OpenStack. But if a host is not publicly accessible, which is the case when using libvirt, it can be configured to obtain a certificate from a certificate authority dedicated to the Enough instance. The default for certificate_authority should be set in ~/.enough/example.com/inventory/group_vars/all/certificate.yml, using this example.

The default can also be changed for a given host (for instance weblate-host) by setting the desired value in the ~/.enough/example.com/inventory/host_vars/weblate-host/network.yml file.

When using a certificate authority dedicated to the Enough instance, each certificate must be manually renewed after a year. For instance, the certificate of website.example.com can be renewed as follows:

$ rm ~/.enough/example.com/certs/website.example.com*
$ enough --domain example.com service create website

The service create command is idempotent: it will notice that the certificate is missing, create a new one, upload it, install it and reload the web server.

Note

The Let’s Encrypt certificates are automatically renewed and do not require manual intervention.

OpenStack Attached volumes

Provisioning

A volume can be created and attached to the host. It can be resized at a later time, when more space is needed. For instance, before creating weblate-host, the desired volume size and name can be set in the ~/.enough/example.com/inventory/host_vars/weblate-host/provision.yml file like so:

---
openstack_volumes:
  - name: weblate-volume
    size: 10

Encrypting and Mounting

The volume can then be encrypted, formatted and mounted by specifying the mount point in the encrypted_device_mount_point variable like so:

---
openstack_volumes:
  - name: weblate-volume
    size: 10
encrypted_device_mount_point: /srv

By default Docker or Snap will be set to reside in the encrypted_device_mount_point directory so that the data it contains is encrypted. It can be disabled with the encrypted_volume_for_docker and encrypted_volume_for_snap variables like so:

---
openstack_volumes:
  - name: weblate-volume
    size: 10
encrypted_device_mount_point: /srv
encrypted_volume_for_docker: false
encrypted_volume_for_snap: false

Resizing

The size of a volume can be increased (but never decreased) by modifying the value from (for instance) 10GB

---
openstack_volumes:
  - name: weblate-volume
    size: 10

to 20GB

---
openstack_volumes:
  - name: weblate-volume
    size: 20

The resize operation is done with the following command (the host will be rebooted). If the volume already has the desired size, the command will do nothing.

$ enough --domain example.com volume resize weblate-host weblate-volume

If the volume is mounted as an encrypted partition, it should then be extended to use the additional disk space. There is no need to unmount the partition.

$ enough --domain example.com ssh weblate-host -- sudo cryptsetup resize --key-file=/etc/cryptsetup/keyfile spare
$ enough --domain example.com ssh weblate-host -- sudo resize2fs /dev/mapper/spare

Access

The SSH public keys found in files matching authorized_keys_globs are installed on every machine.

---
authorized_keys_globs:
  - ssh_keys/dachary.pub
  - ssh_keys/glen.pub

OpenStack backups

Restore a service from a backup

To restore the volume attached to a service from a designated backup:

$ enough --domain example.com openstack volume snapshot list
...
| 6b75f34e | 2020-04-12-cloud-volume | None | available | 100 |
...
$ enough --domain example.com backup restore 2020-04-12-cloud-volume

In this example, the restoration is done as follows:

  • The cloud service is created, if it does not already exist.
  • The machine (cloud-host) attached to the volume (cloud-volume) is stopped. The volume is detached and deleted.
  • A new volume cloud-volume is created from the 2020-04-12-cloud-volume backup and attached to cloud-host.
  • The machine (cloud-host) is restarted.

Create a clone of a service from a backup

It is convenient to create a clone of an existing service based on a backup for:

  • testing and experimenting without disrupting production
  • verify an upgrade won’t loose any data
  • teaching
  • etc.
$ enough --domain example.com openstack volume snapshot list
...
| 6b75f34e | 2020-04-12-cloud-volume | None | available | 100 |
...
$ enough --domain example.com backup restore \
         --target-domain test.d.enough.community \
         2020-04-12-cloud-volume

Once the service is cloned, it will be available at https://cloud.test.d.enough.community. In this example, the cloning is done as follows:

  • A dedicated OpenStack region is used to restore the backup

Note

The OpenStack region where the backup is restored is in the clone section of the ~/.enough/example.com/inventory/group_vars/all/clouds.yml file and it can be modified if the default is not suitable.

  • A volume is created from the 2020-04-12-cloud-volume snapshot
  • The cloud service is created (in the region dedicated to restoring the backup) as well as all the services it depends on, if they do not already exist. Including the DNS server.
  • The test.d.enough.community domain is delegated to the DNS server located in the OpenStack region where the backup was restored so that https://cloud.test.d.enough.community resolves to the newly created cloud service.

It is possible restore the service step by step with the following commands:

$ enough --domain example.com backup clone volume \
         --target-domain test.d.enough.community 2020-07-29-cloud-volume
$ enough --domain test.d.enough.community service create cloud
$ enough --domain test.d.enough.community backup restore 2020-07-29-cloud-volume

Restoring a service that requires a VPN

If the service restored in a clone requires a VPN (that is if it runs on an private IP), a new VPN must be setup before the user can access it.

If the service is cloned with:

$ enough --domain example.com backup restore \
         --target-domain test.d.enough.community \
         2020-04-12-cloud-volume

The credentials to connect to the VPN of the clone are found in the ~/.enough/test.d.enough.community/openvpn directory (for instance ~/.enough/test.d.enough.community/openvpn/loic.tar.gz).

Note

Although the loic.tar.gz file has the same name as in the original, it will connect to a the VPN server in the clone. Care must be taken to not override credentials that existed before the cloning operation.

The subnet of internal network of the clone is hardcoded in .enough/test.d.enough.community/inventory/group_vars/all/internal_network.yml:

Low level commands

The following are not useful if only relying on the service command above. They can however be helpful to run Ansible or OpenStack manually.

Adding hosts

The hosts (OpenStack virtual machines) are created automatically when a service is provided. It is however possible to create a new host or destroy an existing one.

The first step is to edit ~/.enough/example.com/inventory/all.yml and add the name of the new host:

---
all-hosts:
 hosts:
  my-host:
  bind-host:
  forum-host:
  ...

Creating a new host:

enough --domain example.com host create my-host

SSH to a host:

enough --domain example.com ssh my-host

Removing hosts

Every host is known to icinga, bind and wazuh and it should be deleted from these services before being removed.

  • Add the host to the deleted-hosts group in ~/.enough/example.com/inventory/all.yml:
---
deleted-hosts:
  hosts:
    some-host:
  • Run the playbook:
enough --domain example.com playbook
  • Physically delete the host
enough --domain example.com host delete my-host

OpenStack CLI

The openstack CLI can be used as follows:

$ enough --domain example.com openstack -- help

Which is exactly equivalent to:

$ OS_CLIENT_CONFIG_FILE=~/.enough/example.com/inventory/group_vars/all/clouds.yml \
  openstack --os-cloud production help

Playbook CLI

The ansible-playbook CLI can be used as follows:

$ enough --domain example.com playbook -- --limit localhost,icinga-host \
  --private-key ~/.enough/example.com/infrastructure_key \
  ~/.enough/example.com/enough-playbook.yml

It implicitly uses the following inventories (via multiple –inventory options), in order (the last inventory listed has precedence):