Using Rancher-Gen to Dynamically Update Docker Configuration Files

pitrhoThis is a guest post by Alejandro Mesa, Full-Stack Software Engineer and Chief Architect at Pit Rho. Introduction Docker and Rancher have made it far easier to deploy and manage microservice-based applications. A key challenge, however, is managing the configuration of services that depend on other dynamic services. Imagine the following scenario: you have multiple backend containers that run your web application, and a few nginx containers that proxy all requests to the backend containers. Now, there’s a new release of the web application that must be deployed, which means new backend containers need to be built and deployed. After they are deployed, the nginx configuration needs to change to point to the new backend containers. So, what do you do with nginx? Do you change the its configuration, build a new container and deploy it? What if there was a way for you to automatically detect the changes on the backend service and dynamically update nginx? That’s where Rancher-Gen comes into play. Rancher-Gen is a Python utility that listens for service changes in Rancher and renders a user-specified Jinja2 template. This allows a user to generate configuration files for existing services based on those changes. In addition, it provides a mechanism to run a notification command after the template has been rendered. Below is a tutorial that describes how to automatically generate an nginx configuration file for a backend service running the ghost blogging platform.

Tutorial

All configuration files described below can be found under the demo directory in the Rancher-Gen repository. Step 1 -  Deploying the ghost service For simplicity, we’re going to use the official ghost image from Docker hub. So, create a docker-compose.yml file and add the ghost service as follows:

ghost:
  image: ghost
  expose:
    - "2368"

Now, deploy the ghost service using Rancher Compose:

$ rancher-compose -p demo up -d ghost

Step 2  -  Create the nginx image with rancher-gen Here is the Dockerfile used to build the nginx image:

FROM phusion/baseimage:0.9.17
MAINTAINER pitrho

# Step 1 - Install nginx and python
ENV DEBIAN_FRONTEND noninteractive
RUN \
 apt-add-repository -y ppa:nginx/stable && \
 apt-get update && \
 apt-get install -y python-software-properties \
   wget \
   nginx \
   python-dev \
   python-pip \
   libev4 \
   libev-dev \
   expect-dev && \
 rm -rf /var/lib/apt/lists/* && \
 chown -R www-data:www-data /var/lib/nginx && \
apt-get clean

# Step 2 - Install rancher-gen
ENV RANCHER_GEN_VERSION 0.1.2
RUN pip install rancher-gen==$RANCHER_GEN_VERSION

# Step 3 - Define services
RUN mkdir /etc/service/nginx /etc/service/rancher_gen /nginxconf
COPY nginx_run /etc/service/nginx/run
COPY rancher-gen_run /etc/service/rancher_gen/run
COPY default.j2 /nginxconf

# Step 4 - Use baseimage-docker's init system.
CMD ["/sbin/my_init"]

# Step 5 - Expose ports.
EXPOSE 80
EXPOSE 443

Let’s break down the Dockerfile step by step. Steps 1 and 2 are self-explanatory: simply install nginx, python and rancher-gen. Step 3 is where we setup the services that run when the image starts. The first service is nginx, and it runs using the file at /etc/servce/nginx/run. The contents of this file are:

#!/bin/bash
 rancher-gen --host $RANCHER_GEN_HOST \
 --port $RANCHER_GEN_PORT \
 --access-key $RANCHER_GEN_ACCESS_KEY \
 --secret-key $RANCHER_GEN_SECRET_KEY \
 --project-id $RANCHER_GEN_PROJECT_ID \
 $RANCHER_GEN_OPTIONS \
 --notify "service nginx reload" /nginxconf/default.j2 /etc/nginx/sites-available/default

Notice how, after the notify step, we pass two paths, namely /nginxconf/default.j2 and /etc/nginx/sites-available/default. The former is the Jinja2 template and the latter is the output location of the rendered template. Below are the contents of the default.j2 file:

upstream ghost.backend {
{% for container in containers %}
  {% if container['state'] == "running" %}
    server {{container['primaryIpAddress']}}:2368;
  {% endif %}
  {% endfor %}
}

server {
  listen 80;
  server_name ghost_demo;
  location / {
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header HOST $http_host;
    proxy_set_header X-NginX-Proxy true;
    proxy_pass http://ghost.backend;
    proxy_redirect off;
  }
}

Steps 4 and 5 of the Dockerfile set the run command \”/sbin/my_init\” in the image and exposes ports 80 and 443. Now, it’s time to build the image:

$ docker build -t="pitrho/nginx-rancher-gen-demo" .

Step 3 — Create the nginx service and deploy it Now that we have the nginx image, add the nginx service to the docker-compose.yml file created in Step 1:

ghost:
 image: ghost
 expose:
 - "2368"

nginx:
 image: pitrho/nginx-rancher-gen-demo:latest
 ports:
 - 80:80
 links:
 - ghost
 environment:
 NGINX_RUN_TYPE: rancher-gen
 RANCHER_GEN_HOST: $RANCHER_HOST
 RANCHER_GEN_PORT: $RANCHER_PORT
 RANCHER_GEN_ACCESS_KEY: $RANCHER_ACCESS_KEY
 RANCHER_GEN_SECRET_KEY: $RANCHER_SECRET_KEY
 RANCHER_GEN_PROJECT_ID: $RANCHER_GEN_PROJECT_ID
 RANCHER_GEN_OPTIONS: --stack demo --service ghost

The RANCHER_GEN_OPTIONS environment variable above is used to pass additional command-line options to rancher-gen. See the Rancher-Gen documentation for an explanation of these options. Now run rancher-compose to start the nginx service:

$ rancher-compose -p demo up -d nginx

At this point, both the ghost and nginx services should be up and running. ghost nginx check and ghost can be accessed by pointing your browser to the IP address of the host running the nginx container: ghost screen If you were to inspect the nginx container using the shell, and open the rendered file /etc/nginx/sites-enabled/default, you will see the following output:

upstream ghost.backend {
   server 10.42.136.216:2368;
 }

server {
     listen 80;
     server_name ghost_demo;

     location / {
         proxy_set_header X-Real-IP $remote_addr;
         proxy_set_header HOST $http_host;
         proxy_set_header X-NginX-Proxy true;
         proxy_pass http://ghost.backend;
         proxy_redirect off;
     }
  }

As expected, this is the rendered output based on the template specified when running the rancher-gen command. At this point, if you were to upgrade the ghost service, and again look at the rendered file, you would notice that the IP address under the upstream section has changed.

Conclusion

To recap, Rancher-Gen is an automation utility that can be used to generate files and run notification commands. With the expressiveness of Jinja2 templates, and its clean command line interface, Rancher-Gen can be used to generate most configuration files, and automate tasks that otherwise would be tedious and repetitive for most sysadmins and software engineers. If you have any questions or suggestion on how to improve Rancher-Gen, feel free to reach us through the github repository, or contact us on Twitter @PitRho.

快速开启您的Rancher之旅