Code samples from "Dockerizing a ColdFusion Enterprise Application, a Case Study"

ColdFusion Summit 2017, November 16, 2017 1:45 PM - 2:45 PM

These demos were performed on Windows 10 with Docker Version 17.09.0-ce-win33

BINARIES NOT INCLUDED

These files are referenced by the demo code and required for building Docker images but are not included in this distribution:

• ColdFusion_2016_WWEJ_linux64.bin (adobe.com)
• hotfix-004-302561.jar (adobe.com)
• mysql-connector-java-5.1.42-bin (mysql.com)

Demo 1: Minimal Example

A minimal example of building a CF Docker image.

• build the image
• run the container

REQUIREMENTS

• The build requires the Linux x64 ColdFusion installer ColdFusion_2016_WWEJ_linux64.bin file to be in the cfdemo1 folder.

TO BUILD

cd cfdemo1
docker build -t cfdemo1 .

FILES OF INTEREST

Dockerfile
myapp\index.cfm

TO RUN

docker run -ti --rm -p 8500:8500 cfdemo1

Test by browsing to http://localhost:8500

Demo 1: Developer Use Case

Files on the host can be bind mounted to the container using -v c:\host\path:/container/path.

RUN DEVELOPER USE CASE

docker run -ti --rm -p 8500:8500 -v %cd%\myapp:/opt/coldfusion2016/cfusion/wwwroot/myapp cfdemo1

Test by browsing to http://localhost:8500/myapp and modifying files in myapp\.

Base Image (Required for Demo 2)

A base image used by some of the examples.

TO BUILD

cd cfbaseos
docker build -t cfbaseos .

Demo 2: Real-World Example

A real-world example of building a CF Docker image including (everything but the kitchen sink):

• uses a local base image cfbaseos to eliminate 3rd party build-time dependencies
• uses multi-stage builds feature and --squash to reduce the image size
• automates CF hotfix 4 installation
• automates keytool for certificate installation
• relocate the CF web root to /var/www in server.xml
• bind mount the project’s source code www\ to /var/www
• bind mount the CF configuration files cf\ to their respective locations in the CF directory
• add an NGINX service to the stack which will function as a reverse proxy and a static cache
• use docker-compose.yml to define the composition
• use docker-compose up command to bring up the stack

REQUIREMENTS

• The following binaries are required for building and should be put in the cfdemo2\cf-res\software folder:
  • ColdFusion_2016_WWEJ_linux64.bin
  • hotfix-004-302561.jar
  • mysql-connector-java-5.1.42-bin
• The cfbaseos image should be built if it hasn't been already.

TO BUILD

cd cfdemo2
docker build --squash -t cfdemo2 .

FILES OF INTEREST

Dockerfile
cf\ (neo-*.xml, server.xml, seed.properties, etc.)
nginx\nginx.conf
www\
docker-compose.yml

TO RUN

docker-compose up

Test by browsing to http://localhost

Demo 3: Load-balancing with Docker, NGINX, and Redis

Builds on Demo 2 and demonstrates:

• Docker's built-in load-balancer
• builds on the image cfdemo2 built in Demo 2
• adds a simple layer to the CF image: Redis configuration (in neo-runtime.xml)
• adds a Redis service (scale=1) to the composition using the official Redis image
• configures NGINX (scale=1) to look for 3 CF backends
  • note: the upstream backends defined in nginx.conf are based on the way Docker automatically names the containers on the network: cfdemo3_cf_1, cfdemo3_cf_2, etc.
• scales CF using docker-compose --scale
• uses docker-compose.yml to define the composition

REQUIREMENTS

• cfdemo2 image created in Demo 2

TO BUILD

cd cfdemo3
docker build -t cfdemo3 .

FILES OF INTEREST

Dockerfile
cf\
nginx\nginx.conf
www\Application.cfc (enable session mgmt)
docker-compose.yml

TO RUN

docker-compose up --scale cf=3

Test by browsing to http://localhost. Note the containers are load balanced (thanks to Docker) and the session id remains the same (thanks to Redis).

A FEW NOTES ABOUT STICKY SESSIONS INSTEAD OF REDIS

You can't use sticky sessions with Docker load-balancing because the back-ends are non-deterministic. One option is to use NGINX for load-balancing. NGINX can do this by way of the upstream 'ip_hash' directive which by default switches based on client ip. This is a bit tricky though because the client ip gets abstracted by the Docker network NAT. Read more about it here:

docker/for-mac#180

You can alternatively hash on some other client identifier other than client ip like cftoken for example but it takes at least an initial round trip through the load balancer for new clients to receive a cftoken the first time so in these cases the session may not stick on the first request.