capsul-flask/docs/deployment.md

4.2 KiB

Deploying Capsul on a server

Capsul has a "hub and spoke" architecture. The "Hub" runs the web application and talks to the Postrges database, while the "Spoke"s are responsible for creating/managing virtual machines. One instance of the capsul-flask application can run in both hub mode and spoke mode at the same time, however there must only be one instance of the app running in "Hub" mode at any given time.

Installing prerequisites for Spoke Mode

On your spoke (see Architecture You'll need libvirtd, dnsmasq, and qemu-kvm, plus a /tank diectory with some operating system images in it:

sudo apt install libvirt-daemon-system virtinst git dnsmasq qemu qemu-kvm
sudo mkdir -p /var/www /tank/{vm,img,config}
sudo mkdir -p /tank/img/debian/10
cd !$
sudo wget https://cloud.debian.org/images/cloud/buster/20201023-432/debian-10-genericcloud-amd64-20201023-432.qcow2 -O root.img.qcow2

TODO: network set-up

cyberia-cloudinit.yml

The create vm shell script depends on this file. I think right now its file path is hard-coded to /tank/config/cyberia-cloudinit.yml

#cloud-config
preserve_hostname: true
users:
  - name: cyberian
    groups: wheel
    lock_passwd: true
    shell: /bin/bash
    sudo: ['ALL=(ALL) NOPASSWD:ALL']
    ssh-authorized_keys:

Metrics

Capsul integrates with Prometheus to grab the data for the capsul metric graphs.

The code that generates those metrics is here: https://github.com/kumina/libvirt_exporter

autonomic / servers.coop version for ipv6 support

#cloud-config
package_upgrade: true
packages:
  - curl
  - wget
  - git
  - vim
bootcmd:
  - echo 'iface ens3 inet6 dhcp' >> /etc/network/interfaces.d/50-cloud-init
  - ifdown ens3; ifup ens3
final_message: "The system is finally up, after $UPTIME seconds"
users:
  - name: user
    groups: sudo
    sudo: ['ALL=(ALL) NOPASSWD:ALL']
    ssh-authorized-keys:

Deploying capsul-flask

Manually

Follow the local set-up instructions on your server.

Make sure to set BASE_URL correctly, generate your own secret tokens, and configure your own daemon management for the capsul-flask server (e.g. writing init scripts, or SystemD unit files).

Use the suggested gunicorn command (with appropriately-set address and port), instead of flask run, to launch the server.

For example, here is the SystemD service unit file we use in production for capsul.org:

[Unit]
Description=capsul-flask virtual machines as a service
After=network.target

[Service]
ExecStart=/usr/local/bin/pipenv run gunicorn --bind 127.0.0.1:5000 -k gevent --worker-connections 1000 app:app
Restart=on-failure
WorkingDirectory=/opt/capsul-flask

[Install]
WantedBy=multi-user.target

TODO: cron runner is required to run maintenance tasks for now, but in the future we want to build this into the python based task scheduler.

Using Co-op Cloud's vanilla Docker Swarm configuration

Download the Co-op Cloud swarm compose.yml:

wget https://git.autonomic.zone/coop-cloud/capsul/src/branch/main/compose.yml

Optionally, download add-on compose files for Stripe, BTCPay, and Spoke Mode:

wget https://git.autonomic.zone/coop-cloud/capsul/src/branch/main/compose.{stripe,btcpay,spoke}.yml

Then, create a .env file and configure appropriately -- you probably want to define most settings in the Co-op Cloud .envrc.sample file.

Load the environment variables (using Python direnv, or a manual set -a && source .env && set +a), insert any necessary secrets, then run the deployment:

docker stack deploy -c compose.yml -c compose.stripe.yml your_capsul

(where you'd add an extra -c compose.btcpay.yml for each optional compose file you want, and set your_capsul to the "stack name" you want).

TODO: cron runner

Using Co-op Cloud's abra deployment tool

Follow the guide in the README for the Co-op Cloud capsul package.

Using docker-compose

TODO