fail2ban-geo-exporter

It's all in the description already. This little program will expose Prometheus metrics for every IP banned by fail2ban. You can have them geotagged or disable tagging if you just want the IP and the jail it's coming from.

NOTE: This software assumes your fail2ban jail definitions are all in a single file. Multiple definites under jail.d/ are currently not supported.

Disclaimer

This exporter goes deliberately against best practices and ist not suitable for deployments at scale. It's intended to be used in a homelab alike setting and won't even provide any sane metric to alert on. This may change in the future, but it more than likely will not.

By enabling grouping in the conf.yml, the growth of label cardinality can be reduced, but this is still far from ideal.

Metrics

This exporter provides a single time series named fail2ban_banned_ip for each IP in a fail2ban jail. Default labels are: jail and ip with their respective values.

More labels can be provided by enabling geoIP annotation. At this point, the maxmind provider adds city, latitude, and longitude in addition the default labels.

If you enable grouping in the conf.yml, you will instead receive two sets of metrics and no more data about single IPs. Instead you get a gauge fail2ban_location which counts the number of banned IPs per location. The labels are the same as above, just without jail and ip.

The second metric is fail2ban_jailed_ips which is a gauge, that displays all currently banned IPs per jail. jail is the only label in this metric.

It's highly recommended to enable grouping, in order to reduce the cardinality of your labels.

A small guide to creating your own geoIP provider can be found in the Extensibility section of this README.

Configuration

server:
    listen_address:
    port:
geo:
    enabled: True
    provider: 'MaxmindDB'
    enable_grouping: False
    maxmind:
        db_path: '/f2b-exporter/db/GeoLite2-City.mmdb'
f2b:
    conf_path: '/etc/fail2ban'
    db: '/var/lib/fail2ban/fail2ban.sqlite3'

Just plug in the port and IPv4 address you want your exporter to be listening on. If you want to enable geotagging, there is only one method at this time and for that you will need to sign up for a free account at https://www.maxmind.com, download their city database and plug the path to the db in geo.maxmind.db_path. Their paid tier claims to have increased accuracy and is interchangable with their free database, so that should work as a data source for this exporter as well. At the time of writing I can neither deny, nor confirm these claims.

f2b.conf_path assumes default directory structure of fail2ban. So your jails can be defined in /etc/fail2ban/jail.local or in /etc/fail2ban/jail.d/*.local. Default values defined in jail.local (i.e.: bantime) will be picked up and consequently applied to all jails defined under jail.d.

Missing entries in the MaxmindDB will be discarded by default. If you want to keep track of missing entries, you can provide default values to be used instead. In your conf.yml add a section under geo.maxmind like so:

geo:
    maxmind:
        on_error:
            city: 'Atlantis'
            latitude: '0'
            longitude: '0'

Installation

As a systemd service

Pick your favourite working directory and git clone https://github.com/vdcloudcraft/fail2ban-geo-exporter.git .

Now run these commands to set up a python virtual environment:

python -m venv .
. bin/activate
pip install -r requirements.txt

Create a file called fail2ban-geo-exporter.service at /etc/systemd/system/

Open that file and paste following content in there:

[Unit]
Description=fail2ban geo exporter
After=network.target

[Service]
Type=simple
User=root
WorkingDirectory=<path>
Environment=PYTHONPATH=<path>/bin
ExecStart=<path>/bin/python3 <path>/fail2ban-exporter.py
Restart=always

[Install]
WantedBy=multi-user.target

Make sure to replace all four(4) instances of <path> in that config with your actual working directory.

Make sure you have a conf.yml in your working directory as described in Configuration

When that is all done, run following commands and your exporter is running and will survive reboots:

sudo systemctl daemon-reload
sudo systemctl enable fail2ban-geo-exporter.service
sudo service fail2ban-geo-exporter start

You should see a list of metrics when you run curl http://localhost:[port]/metrics

As a Docker container

Make sure you have prepared a config as described in Configuration.

Docker images are provided via Docker Hub

To run the exporter in a Docker container, execute the following command

docker run -d \
        -v /etc/fail2ban:/etc/fail2ban:ro \
        -v /var/lib/fail2ban/fail2ban.sqlite3:/var/lib/fail2ban/fail2ban.sqlite3:ro \
        -v /<path-to-your-db>/GeoLite2-City.mmdb:/f2b-exporter/db/GeoLite2-City.mmdb:ro \
        -v /<path-to-your-conf.yml>/conf.yml:/f2b-exporter/conf.yml \
        --name fail2ban-geo-exporter \
        --restart unless-stopped \
        vdcloudcraft/fail2ban-geo-exporter:latest

Make sure that your paths to jail.local and fail2ban.sqlite3 are correct.

Extensibility

Currently there is only one way to geotag IPs and that is with the Maxmind db. But there is a way to provide custom geoIP providers.

If you wish to implement your own method for annotating your IPs, you need to create a Python class and save the module in ./geoip_provider/

You need to ensure following requirements are fullfilled:

Your module name is a lower case version of your class name. E.g.: Class FancyProvider, module ./geoip_providers/fancyprovider.py
Your class has a constructor that accepts a single parameter. This will be the parsed conf.yml that will be passed to the class.
Your class implements a method annotate(ip), that takes in a single IP as a string and returns a dictionary with all additional labels for Prometheus to use. Do not include the IP itself as a label.
Your class implements a method get_labels() that returns a list of strings with the label names it's going to provide.

When all that is given, you can just put your class name with[!] capitalization into the configuration under geo.provider and enjoy the fruits of your labour.

Be aware that the enable_grouping setting will use only the labels provided by your class to aggregate locations.

If you do create another provider class and think other people might find it useful too, I'll gladly review pull requests.

Grafana dashboard

The files dashboard-*.json include a complete definition for either grouping configuration to display your banned IPs on a worldmap and count all banned IPs per jail.

dmattosr/fail2ban-geo-exporter