I use (and have done for a few years, across versions from 5 forward) a procedure where my implicit context load (for DI jobs) retrieves context variables from a database. So that there is absolutely no difference required from first DEV compilation, through QA to PROD, I configure each of my environments with context database credentials using system variables. This means I can configure each of my environments on creation and do not have to touch my jobs again no matter which environment they run on. I only use the Default context because having multiple contexts will end up with issues no matter how thorough you think you....especially if there is a team of developers. I do this with my ESB work as well, but use PropertyPlaceholder to receive my "context"" variables so that I have exactly the same values across the board. I also add another layer of complexity in many environments as many of my customers do not like human readable values to be stored even in environment settings. For this I use JASYPT. I do not experience this issue unless service users are incorrect OR there is a new Job Server that has not been configured. I have used this on Windows and Unix and apart from the configuration of the servers, the implementation is exactly the same.
I use this routine code in my Implicit Context settings (they can run code as well as accept hardcoded values) and within my jobs sometimes....although I tend to load everything from the Implicit Context if it is suitable.
/**
* getEnvironmentVariable: used to retrieve Environment Variables
*
* {talendTypes} String
*
* {Category} Implicit Context Load
*
* {param} string("TalendContextPassword") variableName: the parameter name to be returned
*
* {example} getEnvironmentVariable("TalendContextPassword") # returns "My Password"
*/
public static String getEnvironmentVariable(String variableName) {
String returnVal = System.getenv(variableName);
if (returnVal == null) {
System.out.println(variableName
+ " does not exist or holds no value");
}
return returnVal;
}
This works for running on my DEV Studio (my machine is configured with environment variables on Windows), running jobs from the command line (Windows or Linux) and running by starting on TAC. To add even more of a potential for problems to this mix, I automatically load tasks into the TAC using my ESB runtime (when a file is dropped in a folder and the required job is not yet live) by using the MetaServlet API to find the required job in my Nexus and load to the TAC. No contexts are set up in the TAC because this process renders doing that unnecessary. It also enables me to have new jobs that supersede already existing jobs to be used as soon as they are logged in my background database as being the "live" job.
This all works without fault and n my current project I have had no problems with this across 3 environments for the last 6 months. As I said, there is nothing in Talend stopping this from working. It is an implementation problem.
