Clooster is a Perl script designed to automatically switch the address of a DNS record whenever a server goes offline.
Clooster uses a client-server architecture, which removes the need of an additional monitoring server. It also has the advantage of being expandable, even if right now it only supports a maximum of two servers.
- Extremely easy to use and to configure: you need two servers, a Cloudflare account with an host, a JSON file and a Perl interpreter.
- Cloudflare and Pushbullet support out of the box, no extra modules needed.
- Small memory footprint and low number of dependencies. The script itself is only ~10KB!
- Depending on the kind of failure, Clooster detects an offline server instantly or in a few minutes.
- Only one script for both the client and the server.
Beware! This script has been written for fun, and I do not have the time and resources required to test it extensively. Don't blame me if something explodes, or if it doesn't work at all!
Instead, why don't you write a an issue or a pull request? ♥
Clooster depends on:
- Perl 5.10.1 or better.
- Mojolicious, a lightweight framework with a lot of cool stuff.
- JSON::MaybeXS, a module which does the right thing when decoding JSON.
- Create a
CNAMErecord on your domain pointing to your preferred host. - Get an API key from the account page of Cloudflare.
- Optionally, get an API key from Pushbullet.
- Configure the server and client instances of Clooster (you need two separate configs).
See
clooster.json.example. - Put Clooster on your servers, and run the two instances (first the server one, then the client one). Don't forget to install any missing dependency with cpan.
The path of the config file can be specified with the first argument passed to Clooster.
It defaults to %{script_name}.json.
It's more fun when you do it by yourself.
Well, I was thinking about cluster and max two servers (2), so this name came out of my mind.
I'm not good at picking names.
Since version 1.1, an example fail2ban
configuration is provided.
The configuration enables automatic IP banning (by default, for 1 hour) of unknown clients that
try to connect to the clooster server instance. Two attempts are sufficient for a ban to occur.
For this to work, Clooster needs to be started by systemd, and its output has to be recorded
by the journal (the default). See How do I daemonize the script?
NOTE: before using the included files, you need to change the port Clooster listens on in
fail2ban/jail.d/10_clooster.conf.
It's actually pretty simple. First and foremost, the script calls Cloudflare and gets the zone id and an object representing the record that will be updated. The server then performs a DNS lookup on the hostname of the other server, and saves the retrieved IP addresses for later. The server binds on the address specified in the config and lets the client connect. Only one client is allowed at a time. Other connections are instantly terminated.
When the client successfully connects to the server, its IP is checked by the server: if it does not exist in the list of IPs saved before, then the client is promptly rejected. The authorization string is now sent: it's simply the server name encoded with HMAC-SHA256 (with the key specified in the configuration). Nothing too fancy, it's just one more way to ensure that the script is talking to the right server.
Once the authorization is complete, the script sends a keep-alive string every 60 seconds to keep
the socket alive. The keep-alive string is the server name (this_server in the config).
A watchdog runs every 120 seconds, and does the following:
- Ensures that the client is connected to the server, and if it is not, a connection is attempted.
- Ensures that the keep-alive strings are being sent, by comparing the timestamp of the last keep-alive and the current system time.
When the socket is closed (either after a timeout, or simply because one server died), the
down_handler() function of the script is called. An up-to-date version of the record object is
retrieved from Cloudflare, to prevent wrong assumptions derived from old data.
Once the updated record is available (and even if it isn't - errors are suppressed in this case),
the script checks if the record is already set to the value of this_server in the config.
If it is, then there is nothing to do - it just notifies the administrator using Pushbullet, if
enabled, and then waits until the other server is up again. Otherwise, the record is updated
remotely by using the Cloudflare API, and a notification is sent.
Once the dead server is up again (and with it, the script), the same thing as before is performed,
except that the record is changed to the value of cloudflare.preferred_value in the config.
You can write an init script for your system init daemon. Systemd can do this and much more.
An example unit file for systemd is provided with the script. See clooster.service.
You have to use a tool like curl from the command line to perform the task.
curl -H 'Authorization: Bearer <access_token>' -X GET https://api.pushbullet.com/v2/devicesWhy did you bother writing modules wrapping the API of Cloudflare and Pushbullet, when they are already available on CPAN?
Two reasons:
- I designed the script to be as lightweight as possible in terms of dependencies and memory footprint.
- I also wanted to perform as many operations as possible asynchronously, and the easiest way was to use Mojo::UserAgent.
Also, keep in mind that my implementations are specific to the script! They do not implement all the methods of the respective APIs, and include a few quirks I am not proud about.
This is some of the stuff I'd add in another life:
add a version number :Pdone!- command-line interface (with
Getopt::Long) - conversion to the standard Perl application interface (
App::cloostermaybe?) - support for multiple servers
- modularity (support other DNS services and notification methods)
Copyright (C) 2015, Roberto Frenna.
This program is free software, you can redistribute it and/or modify it under the terms of the Artistic License version 2.0.