Manual installation¶
User attention is required
Kustosz is flexible and supports deployments to various environments. This guide limits a number of choices covered to lower cognitive load and provide instructions that are easier to follow. However, it still expects you to understand what is happening and adjust specific steps to your particular environment. It is recommended that you familiarize yourself with entire document before taking any action, including Additional setup instructions section.
This page covers manual installation of Kustosz. It’s intended for users who require maximum flexibility or want deeper understanding of Kustosz deployments. Majority of users should be able to deploy Kustosz using automatic installer.
Prerequisites¶
Kustosz is web application. You probably want it available under a dedicated (sub)domain that supports secure HTTPS connection. Configuration of domain, HTTPS and access control is outside of scope of this guide.
There are few dependencies that you absolutely must meet:
Python 3.9 or newer
Ability to execute arbitrary commands in the context of server
Ability to start your own long-running processes (needed by Celery worker)
At least 256 MiB of memory
While not required, we also recommend:
PostgreSQL database server
Redis as Celery transport
Prepare Kustosz server environment¶
Create Kustosz base directory. This is where you will put settings, and where Kustosz will store SQLite database file and certain cache files.
mkdir ~/kustosz_home
Create environment variables required by Kustosz. You need to ensure they are set for all processes that involve Kustosz, e.g. Celery or cron job. You might want to add them to ~/.bash_profile
.
export ENV_FOR_DYNACONF="production"
export DJANGO_SETTINGS_MODULE=kustosz.settings
export KUSTOSZ_BASE_DIR="$HOME/kustosz_home"
At this stage, do all the other setup that your Kustosz installation might require - create (sub)domain, create Postgres user and database, install Redis server and ensure it starts automatically etc. The exact actions to take depend on your operating system and preferred configuration.
Install Kustosz server¶
The very first step is creating and activating virtual environment. It must exists as long as you run Kustosz, so you must create it in a path that won’t disappear automatically.
You might need to re-create virtual environment once your system Python is upgraded.
It’s fine to remove virtual environment and create new one when you upgrade Kustosz to newer version.
python3 -m venv ~/.virtualenvs/kustosz/
source ~/.virtualenvs/kustosz/bin/activate
Activating virtual environment with environment variables
Instead of sourcing activation script, you may also set two environment variables:
export VIRTUAL_ENV="$HOME/.virtualenvs/kustosz"
export PATH="$VIRTUAL_ENV/bin:$PATH"
Once virtual environment is created and activated, you can install Kustosz package:
pip install --upgrade pip wheel
pip install kustosz
(Optional) Install additional Python packages¶
Depending on your preferred configuration, you might need to install additional Python modules, such as:
redis
is required to use Redis for cache or as Celery transportpsycopg2
is required to connect to PostgreSQL database (psycopg2 requires additional system packages to build)pylibmc
orpymemcache
is required to use Memcached for cachewhitenoise
may be used to serve static files directly from WSGI server
pip install redis psycopg2
(Optional) Install kustosz-node-readability
¶
While syndication formats exists so users can read content in their reader applications, many authors try to force readers to visit the website. Kustosz can automatically extract full article content from source page. This extraction is done by “readability” algorithm.
If you have Node.js on your server, you can install kustosz-node-readability. It uses Readability.js, algorithm implementation used for Firefox Reader View and by Pocket. It’s probably the best maintained readability implementation around.
npm install -g kustosz-node-readability
Configure Kustosz¶
The main configuration file is $KUSTOSZ_BASE_DIR/settings/settings.yaml
. You can download sample file from git repository.
mkdir -p $KUSTOSZ_BASE_DIR/settings/
curl https://raw.githubusercontent.com/KustoszApp/server/main/settings/settings.yaml -o $KUSTOSZ_BASE_DIR/settings/settings.yaml
You can modify sample file, or put site-specific modifications in $KUSTOSZ_BASE_DIR/settings/settings.local.yaml
. Second option is preferred, as it allows you to overwrite settings.yaml
file during upgrade, while keeping site-specific settings safely separated.
The exact changes to make are highly dependant on your preferred configuration. Make sure to also read backend configuration page.
Below is non-exhaustive list of settings you might want to change. Consult Kustosz, Django and Celery settings documentation for meaning of specific options.
SECRET_KEY
should be set to new value, obtained by runningkustosz-manager generate_secret_key
ALLOWED_HOSTS
should contain fully qualified domain name where Kustosz is runningDATABASES
, especially if you want to use PostgreSQLCACHES
, if you want to use Memcached or Redis for cacheCORS_ALLOWED_ORIGINS
, if you decide to run frontend and backend on separate domains, this should contain origin of frontend pageCELERY_BROKER_URL
, if you decide to use Redis as Celery brokerKUSTOSZ_READABILITY_NODE_ENABLED
, if you have installed kustosz-node-readabilitySTATIC_ROOT
, if you decide to serve static files using gunicorn; this is path to directory where files will be copied to, it should be inside$KUSTOSZ_BASE_DIR
STATICFILES_DIRS
, if you decide to serve static files using gunicorn; this is list of paths of static files to copy, it should contain directory where you extract frontend files
Example configuration
Settings below assume that you serve Kustosz from kustosz.example.com
, use PostgreSQL as database and Redis as Celery transport.
production:
DATABASES:
default:
ENGINE: 'django.db.backends.postgresql_psycopg2'
NAME: "kustoszdb"
USER: "postgres"
PASSWORD: "postgres_user_password"
HOST: "localhost"
PORT: "5432"
CELERY_BROKER_URL: "redis://localhost:6379/"
ALLOWED_HOSTS:
- 'kustosz.example.com'
STATIC_ROOT: '/home/USER/kustosz_home/frontend_assets'
STATICFILES_DIRS:
- '/home/USER/kustosz_frontend'
Run migrations¶
Migrations change database structure. You have to run migrations before you first start Kustosz, and each time after upgrading to new version.
kustosz-manager migrate
Create cache tables¶
In default configuration, Kustosz uses main database for cache. Database must first be prepared for that. You need to run this only once, before you first start Kustosz. You don’t have to run it if you use Memcached or Redis for cache.
kustosz-manager createcachetable
Start main Celery process¶
Kustosz requires main Celery process to be running. It is used to handle all background tasks, such as downloading files or running filters.
Command to start Celery process is:
celery -A kustosz worker -l INFO -Q feed_fetcher,celery
Due to way Kustosz is designed, some background tasks should run serially. The best way of ensuring that is by running separate Celery process with single worker for feed_fetcher
queue. The downside is that it requires more system resources.
celery -A kustosz worker -l INFO -Q celery
celery -A kustosz worker -l INFO -Q feed_fetcher --concurrency 1
It is recommended that you use process supervisor to run commands above. Most Linux systems come with systemd, which may be used for that purpose - see Use systemd to ensure background processes are running. Another option is Supervisor - see Use supervisor to ensure background processes are running.
Run gunicorn¶
It’s finally time to start main Kustosz server. We install and run gunicorn, but you can use any WSGI-compatible web server. uWSGI is another popular option. If you use Apache web server, you may want to use mod_wsgi.
pip install gunicorn
gunicorn kustosz.wsgi --bind 0.0.0.0:8000
Download frontend files¶
Kustosz server provides REST API only. To make use of it, you also need a client.
Create new directory on your server. Open Kustosz web UI releases page in your browser and find kustosz.tar.xz
file provided by the newest release. Download that file into created directory and unpack it. You can remove archive afterwards.
mkdir ~/kustosz_frontend
cd ~/kustosz_frontend
curl -L https://github.com/KustoszApp/web-ui/releases/download/VERSION_NUMBER/kustosz.tar.xz -o kustosz.tar.xz # change VERSION_NUMBER to latest version number
tar xf kustosz.tar.xz
rm kustosz.tar.xz
Set up HTTP proxy in front of gunicorn¶
Most of dedicated WSGI web servers consider serving static files as out of their scope. Usually it is recommended that you set up HTTP proxy in front of WSGI server. NGINX is popular web server that is commonly used as such proxy. Most Linux distributions have NGINX package that you can install.
After installing NGINX, create /etc/nginx/sites-available/kustosz
file with following content:
upstream kustosz {
# port number here must be the same as port number in gunicorn --bind command
server localhost:8000;
}
server {
listen 80;
# domain name that Kustosz will be available at
server_name localhost;
location / {
proxy_pass http://kustosz;
proxy_set_header Host $host;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_redirect off;
client_max_body_size 10M;
}
location /ui/ {
# this must be absolute path to directory where you extracted frontend files
alias /home/USER/kustosz_front/;
}
}
Make configuration available to NGINX server and restart it so changes can be applied:
ln -s /etc/nginx/sites-available/kustosz /etc/nginx/sites-enabled/
systemctl restart nginx
NGINX file permissions
NGINX often runs as an user with limited permissions, such as www
, www-data
, nginx
or nobody
. You need to ensure that this user has permissions to read extracted frontend files. It also needs executable bit on all directories in the path.
Congratulations! You should see Kustosz running at http://your-domain/ui/.
Additional setup instructions¶
Following section documents configuration changes that do not apply to all deployments, but are still common. Some of these instructions extend basic installation described above, while some override parts of it.
Set up periodic channels update with Celery¶
As noted in automatic channels update section, Kustosz requires channels update process to run periodically. The preferred way of ensuring this process is run is by using Celery beat. Celery beat must run in the background all the time in addition to main Celery process.
celery -A kustosz beat -l INFO
Set up periodic channel update with cron¶
As noted in automatic channels update section, Kustosz requires channels update process to run periodically. If you can’t run Celery beat, you may use system scheduler, such as cron.
The command that you want to run is:
kustosz-manager fetch_new_content --wait
To run it with cron, use crontab -e
command and add the following line:
*/5 * * * * kustosz-manager fetch_new_content --wait
cron jobs usually run in special environment that differs from normal shell. Remember that Kustosz command line tools require few environment variables and active virtual environment. You might want to create simple wrapper script that sets up these variables before running the command. Example of such wrapper is available in source code repository in etc/cron/kustosz-fetch-content.sh
.
Use systemd to ensure background processes are running¶
Kustosz requires some background processes to run all the time - main Celery process, Celery beat process (optionally), gunicorn WSGI server (optionally). One way to ensure they are running is systemd. Unlike supervisor, systemd can start processes automatically during system boot.
Systemd is default init in most of Linux distributions.
Sample systemd unit file is available in source code repository in /etc/systemd/system/kustosz@.service
. This is unit template that calls dispatcher script to start the actual process. Dispatcher script is responsible for setting required environment variables and activating virtual environment. Sample dispatcher script is provided in /etc/systemd/bin/kustosz-service-dispatcher
.
Put dispatcher script anywhere on your file system and unit file in /etc/systemd/system/
. Start and enable all background processes:
systemctl enable --now kustosz@worker.service
systemctl enable --now kustosz@feedfetcher.service
systemctl enable --now kustosz@clock.service
systemctl enable --now kustosz@web.service
Use supervisor to ensure background processes are running¶
Kustosz requires some background processes to run all the time - main Celery process, Celery beat process (optionally), gunicorn WSGI server (optionally). One way to ensure they are running is supervisor.
You can install supervisor in your virtual environment with pip install supervisor
.
Sample supervisor config is available in source code repository in etc/supervisor/supervisord.conf
. Feel free to adjust it to your liking and to better fit your environment. Sample config file requires run
and logs
directories to exist in the same directory as supervisord.conf file and does not start gunicorn process automatically.
Finally, start supervisor:
supervisord -c path/to/supervisord.conf
supervisor can ensure that certain processes are running, but it has one drawback - you need to start it manually. supervisor documentation has section on starting supervisord on system startup. Some Linux distributions, like Debian, offer supervisor package with init daemon script that can be used to start it automatically. If you go down this path, remember that Kustosz command line tools require few environment variables and active virtual environment. You might also need to ensure correct file permissions, as supervisor will be running as privileged user, while Kustosz was installed for standard user.
Use gunicorn to serve static files¶
WhiteNoise is project dedicated to making WSGI-compatible web servers better at serving static files. Serving static content is main reason this guide uses NGINX, so with WhiteNoise you might not need NGINX anymore. The second reason is security, so you probably still want some kind of HTTP proxy in front of WhiteNoise-enabled WSGI server.
First, install WhiteNoise:
pip install whitenoise
Configure Kustosz:
STATICFILES_DIRS
should contain directory where you extracted frontend files; this setting is list of pathsSTATIC_ROOT
should be a path to directory where static files will be copied to; it should be inside$KUSTOSZ_BASE_DIR
STATICFILES_STORAGE
should be set to"whitenoise.storage.CompressedManifestStaticFilesStorage"
MIDDLEWARE
should contain"whitenoise.middleware.WhiteNoiseMiddleware"
as third entry on the list, below"django.middleware.security.SecurityMiddleware"
and"corsheaders.middleware.CorsMiddleware"
; see containersettings.local.yaml
in source code repositorySTATIC_URL
must be set
Finally, run collectstatic
command. This command will walk through all directories specified in STATICFILES_DIRS
and copy files into STATIC_ROOT
, where they will be processed (i.e. compressed).
kustosz-manager collectstatic
You have to run above command before you first start Kustosz, and each time after upgrading to new version.