Deploying Dancer Apps

#cpan #daemon #dancer2

Over the last week or so, as a background task, I’ve been moving domains from an old server to a newer and rather cheaper server. As part of this work, I’ve been standardising the way I deploy web apps on the new server and I thought it might be interesting to share the approach I’m using and talking about a couple of CPAN modules that are making my life easier.

As an example, let’s take my Klortho app. It dispenses useful (but random) programming advice. It’s a Dancer2 app that I wrote many years ago and have been lightly poking at occasionally since then. The code is on GitHub and it’s currently running at klortho.perlhacks.com. It’s a simple app that doesn’t need a database, a cache or anything other than the Perl code.

Dancer apps are all built on PSGI, so they have all of the deployment flexibility you get with any PSGI app. You can take exactly the same code and run it as a CGI program, a mod_perl handler, a FastCGI program or as a stand-alone service running behind a proxy server. That last option is my favourite, so that’s what I’ll be talking about here.

Starting a service daemon for a PSGI app is simple enough – just running “plackup app.psgi” is all you really need. But you probably won’t get a particularly useful service daemon out of that. For example, you’ll probably get a non-forking server that will only respond to a single request at a time. It’ll be good enough for testing, but you’ll want something more robust for production. So you’ll want to tell “plackup” to use Starman or something like that. And you’ll want other options to tell the service which port to run on. You’ll end up with a quite complex start-up command line to start the server. So, if you’re anything like me, you’ll put that all in a script which gets added to the code repo.

But it’s still all a bit amateur. Linux has a flexible and sophisticated framework for starting and stopping service daemons. We should probably look into using that instead. And that’s where my first module recommendation comes into play – Daemon::Control. Daemon::Control makes it easy to create service daemon control scripts that fit in with the standard Linux way of doing things. For example, my Klortho repo contains a file called klortho_service which looks like this:

#!/usr/bin/env perl

use warnings;
use strict;
use Daemon::Control;

use ENV::Util -load_dotenv;

use Cwd qw(abs_path);
use File::Basename;

Daemon::Control->new({
  name => ucfirst lc $ENV{KLORTHO_APP_NAME},
  lsb_start => '$syslog $remote_fs',
  lsb_stop => '$syslog',
  lsb_sdesc => 'Advice from Klortho',
  lsb_desc => 'Klortho knows programming. Listen to Klortho',
  path => abs_path($0),

  program => '/usr/bin/starman',
  program_args => [ '--workers', 10, '-l', ":$ENV{KLORTHO_APP_PORT}",
                    dirname(abs_path($0)) . '/app.psgi' ],

  user => $ENV{KLORTHO_OWNER},
  group => $ENV{KLORTHO_GROUP},

  pid_file => "/var/run/$ENV{KLORTHO_APP_NAME}.pid",
  stderr_file => "$ENV{KLORTHO_LOG_DIR}/error.log",
  stdout_file => "$ENV{KLORTHO_LOG_DIR}/output.log",

  fork => 2,
})->run;

This code takes my hacked-together service start script and raises it to another level. We now have a program that works the same way as other daemon control programs like “apachectl” that you might have used. It takes command line arguments, so you can start and stop the service (with “klortho_service start”, “klortho_service stop” and “klortho_service restart”) and query whether or not the service is running with “klortho_service status”. There are several other options, which you can see with “klortho_service status”. Notice that it also writes the daemon’s output (including errors) to files under the standard Linux logs directory. Redirecting those to a more modern logging system is a task for another day.

Actually, thinking about it, this is all like the old “System V” service management system. I should see if there’s a replacement that works with “systemd” instead.

And if you look at line 7 in the code above, you’ll see the other CPAN module that’s currently making my life a lot easier – ENV::Util. This is a module that makes it easy to work with “dotenv” files. If you haven’t come across “dotenv” files, here’s a brief explanation – they’re files that are tied to your deployment environments (development, staging, production, etc.) and they contain definitions of environment variables which are used to control how your software acts in the different environments. For example, you’ll almost certainly want to connect to a different database instance in your different environments, so you would have a different “dotenv” file in each environment which defines the connection parameters for the appropriate database in that environment. As you need different values in different environments (and, also, because you’ll probably want sensitive information like passwords in the file) you don’t want to store your “dotenv” files in your source code control. But it’s common to add a file (called something like “.env.sample”) which contains a list of the required environment variables along with sample values.

My Klortho program doesn’t have a database. But it does need a few environment variables. Here’s its “.env.sample” file:

export KLORTHO_APP_NAME=klortho
export KLORTHO_OWNER=someone
export KLORTHO_GROUP=somegroup
export KLORTHO_LOG_DIR=/var/log/$KLORTHO_APP_NAME
export KLORTHO_APP_PORT=9999

And near the top of my service daemon control program, you’ll see the line:

use ENV::Util -load_dotenv;

That looks to see if there’s a “.env” file in the current directory and, if it finds one, it is loaded and the contents are inserted in the “%ENV” hash – from where they can be accessed by the rest of the code.

There’s one piece of the process missing. It’s nothing clever. I just need to generate a configuration file so the proxy server (I use “nginx”) reroutes requests to klortho.perlhacks.com so that they’re processed by the daemon running on whatever port is configured in “KLORTHO_APP_PORT”. But “nginx” configuration is pretty well-understood and I’ll leave that as an exercise for the reader (but feel free to get in touch if you need any help).

So that’s how it works. I have about half a dozen Dancer2 apps running on my new server using this layout. And knowing that I have standardised service daemon control scripts and “dotenv” files makes looking after them all far easier.

And before anyone mentions it, yes, I should rewrite them so they’re all Docker images. That’s a work in progress. And I should run them on some serverless system. I know my systems aren't completely up to date. But we’re getting there.

If you have any suggestions for improvement, please let me know.

The post Deploying Dancer Apps first appeared on Perl Hacks.