April 24, 2017
Ideally, application development is separated into three environments: local (development), staging, and production. Staging and production would reside on the IBM i, while local is specific to each developer and would reside on the developer’s local development machine. In a perfect world, the developer would have a mock of the database on their local for development, but DB2 is proprietary to IBM i. Therefore, having a local copy of the database is improbable at best, so how do we gain access to the data for development?
Currently, to my knowledge, there is no way to remotely access DB2 data unless one purchases a license for IBM DB2 Connect. This makes local, staging, and production environment development difficult.
One way to alleviate this frustration is to completely decouple the backend of applications from the frontend by developing a REST, CRUD, or other type of API in order to have web-callable services. This at least allows one to develop the frontend of applications locally. However, then the question arises: what about the API? How does one properly separate the environments for applications that must be developed on the IBM i? What are good practices to keep each developer from stepping on others’ toes?
This article will discuss how to setup a friendly development environment on the IBM i that allows for individual development environments along with staging and production environments. Basic knowledge of git and remote repositories is assumed to keep this light.
File Structure and Apache Configuration
The file structure is important here. It will help us stay organized and give each developer their own environment.
Protip: Keep apache configurations within each specific project. In this example they’re under
project-root/conf. This allows for apache configurations for each environment to be version controlled. It also helps with organization and will make enabling/disabling web applications a breeze. Simply add
IncludeOptional /www/.conf-enabled/*.confto the end of the root Apache configuration file. This is usually found through either the Apache administration web interface or through the IFS. The file is usually located at
/www/zendsvr6/conf/httpd.conf. I leave the
/www/zendsvr6up as a placeholder website to fall back on. It usually just displays something like, “IT WORKS!!!” just to prove apache is at least working.
/ www/ .conf-enabled/ <=== Symlinked apache configuration files for each environment app-one-dev-developer-one.conf <=== Symlinked to /www/development/developer-one/app-one/config/developer-one.conf app-one-stage.conf app-one-production.conf development/ developer-one/ app-one/ <=== Git repo on any development branch conf/ <=== Project-specific, versioned apache configuration files that developer-one.conf developer-two.conf stage.conf production.conf public/ <=== Public web files - some name it htdocs app-two/ <=== Git repo on any development branch app-three/ <=== Git repo on any development branch developer-two/ app-one/ <=== Git repo on any development branch app-four/ <=== Git repo on any development branch stage/ app-one/ <=== Git repo on the stage branch app-two/ <=== Git repo on the stage branch app-three/ <=== Git repo on the stage branch app-four/ <=== Git repo on the stage branch production/ app-one/ <=== Git repo on the master branch app-two/ <=== Git repo on the master branch zendsvr6/ conf/ <=== Root apache configuration files htdocs/ index.html <=== Simple landing page to show apache works. Nothing more.
While the developer could edit the remote files directly, I’ve found it better to sync the files remotely with local files. This gives one a few different benefits. One of which is being able to quickly search the entire project since it is searching local files.
Webstorm, PHPStorm, Netbeans, and a few other IDEs have this capability built in. If one is using Sublime or Atom, I suggest finding a remote syncing plugin they like. I personally like Sublime SFTP.
Doing this is only useful for the development environment folders. Files under stage and production should never
be directly edited. They should only ever need to have
git pull performed on them to sync files.
No DB2 Connect
How does this help us remove the need for DB2 Connect? Well, now we are able to develop all our applications on the IBM i while still retaining development environments as if we were developing locally. This gives us complete access to the DB2 data without the need to access it remotely.
While all applications can be developed with remote syncing, I still suggest decoupling the backend. Not only is this best practice, but it will allow the data to be accessed the same way by many different applications. This helps with consistency, standards, and encapsulation. This will also allow the majority of applications to be developed locally. Ideally there would be one application that need be edited remotely, and that would be the API application. The rest would be frontend applications that consume the API services to deliver fast, user-friendly experiences.