Hosting

Nick edited this page Aug 24, 2015 · 18 revisions

Let's see how would you run thumbor in different environments.

Development Environment

For running it locally you just need to get a proper Configuration file. You can put it at /etc/thumbor.conf, ~/thumbor.conf (home folder) or specify it when starting thumbor.

To verify if you have thumbor, just type:

thumbor --version

It should return the version you've installed. Starting thumbor is as easy as:

thumbor

For more options check the Configuration page.

Production Environment

Other than having the proper Configuration file for your environment, we have some recommendations on how to run thumbor in production.

Our first recommendation is to run more than one instance of it. You can specify different ports using thumbor easily. This will make sure that your service stays responsive even if one of the processes die.

We also recommend having some form of load balance that distributes the load between the aforementioned processes. We are using NGINX to do it, but there are more sophisticated load balance softwares around. thumbor supports health checking under the /healthcheck URI if you need to use it.

Other than that, you run it using the thumbor console app specifying the arguments, like this:

thumbor --port=8888 --conf="~/mythumbor.conf"

We recommend using an application such as Supervisor (http://supervisord.org/index.html) to monitor your services. An example of a supervisord.conf file would be:

[supervisord]
logfile = /home/thumbor/logs/supervisord.log
logfile_maxbytes = 50MB
logfile_backups=10
loglevel = info
pidfile = /home/thumbor/supervisord.pid
user = thumbor

[program:thumbor]
command=thumbor --port=800%(process_num)s --conf=/etc/thumbor800%(process_num)s.conf
process_name=thumbor800%(process_num)s
numprocs=4
user=thumbor
directory=/home/thumbor/
autostart=true
autorestart=true
startretries=3
stopsignal=TERM
stdout_logfile=/home/thumbor/logs/thumbor800%(process_num)s.stdout.log
stdout_logfile_maxbytes=1MB
stdout_logfile_backups=10
stderr_logfile=/home/thumbor/logs/thumbor800%(process_num)s.stderr.log
stderr_logfile_maxbytes=1MB
stderr_logfile_backups=10

This configuration file makes sure that supervisor starts 4 processes of thumbor on the 8000, 8001, 8002 and 8003 ports, each with a different configuration file (thumbor8000.conf, thumbor8001.conf, thumbor8002.conf, thumbor8003.conf all under /etc folder). The other settings are optional, but if you need help with supervisor's settings it has extensive documentation online (http://supervisord.org/introduction.html).

Thumbor in the Cloud

UPDATE

Now there's a project to help with hosting in HEROKU called thumbor-heroku.

Creating your thumbor install in heroku

You can deploy and test Thumbor in the cloud. It's quite easy with Heroku :

  • Create an account like described at http://devcenter.heroku.com/articles/quickstart
  • Install the heroku Toolbelt as described in the same page
  • Log to Heroku in your shell
  • Create a small git project for the configuration of your Thumbor instance.

The whole script to deploy and start an instance :

mkdir heroku
cd heroku/
echo "thumbor>=2.7.0" >> requirements.txt # let heroku deploy and compile prerequisite package via PIP
echo "web: thumbor -p $PORT" >> Procfile # listening port is automatically affected at deployment (we use here the default config)
git init
git add .
git commit -m "init"
heroku create --stack cedar
git push heroku master

Basically, adding thumbor in requirements.txt will install everything you need on Heroku, and you just need to run thumbor -p $PORT to run thumbor on Heroku. In order to run process on Heroku, you need to write down the command in Procfile. Procfile looks like following (make sure there are no "" inside both files):

$ cat Procfile
web: thumbor -p $PORT

Your heroku folder (or whatever you named, I named it thumbor) should look like following (only contains two files):

~/thumbor(master)$ ls
Procfile        requirements.txt
  • Start the instance (Remember: 1 heroku web instance is free of charges, so don't try with more yet):

    heroku scale web=1

  • Verify your new instance is up (in the case of our sample project is stormy-stone-5336.herokuapp.com):

    heroku ps

  • Now if you point your browser to the server name, you'll get a 404 HTTP Error. Just try with an URL that thumbor understands. To open your web browser pointing to the new server:

    heroku open

  • Then try something like:

http://stormy-stone-5336.herokuapp.com/unsafe/300x200/http://s.glbimg.com/jo/g1/f/original/2012/03/16/supersonic-skydiver_fran.jpg

(notice there is no listening port specified)

If you need to scale thumbor server, read more about it in Heroku's documentation.

The sample implementation for the above links can be found at https://github.com/heynemann/thumbor-heroku and is open-source and MIT Licensed.

Another Thumbor/Heroku configuration

This blog post and the attached repositories (Jetpack and thumbor-heroku) explain a more advanced Heroku deployment, that support the smart URL feature.

Thumbor on OpenShift

There's a project showing how to deploy a working version on OpenShift https://github.com/rafaelcaricio/thumbor-openshift-example

Thumbor behind CloudFront

The awesome people at yipit are using thumbor behind the CloudFront CDN at Amazon.

The detailed information on how to do it can be seen at this blog post.

Thumbor with Ansible

ansible-thumbor is an ansible role for installing and configuring thumbor.

Thumbor on AWS OpsWorks

chef-thumbor is a Chef cookbook to setup an entire thumbor stack including NginX, supervisord and mmonit. A quick tutorial how to setup thumbor on AWS OpsWorks can be found here: "thumbor Stack on AWS OpsWorks in 10 minutes".

Thumbor in Docker

There are many Thumbor repos on dockerhub

Getting your own server up and running could be as easy as:

docker run -d -P -e "DETECTORS=['thumbor.detectors.face_detector','thumbor.detectors.feature_detector',]" --name thumbor apsl/thumbor

(that particular image, apsl/thumbor, includes opencv for feature & face detection, nginx cache, s3 support, and circus for process control)