using json instead of a string as context in cdk.json

❓ General Issue

The Question

Hello, I’m trying to setup one cdk,json which will provide parameters for all DEV, QA and PROD environments. My cdk.json looks like this:

"context": {
   "appName": "app1",
   "ENV:TEST": {
          "vpcId": "vpc-111111111111"
    },
   "ENV:QA": {
          "vpcId": "vpc-222222222222"
    },
   "ENV:PROD": {
          "vpcId": "vpc-333333333333"
    }
}

I have it working, but every time i run “cdk synth”, all “ENV:XXX” entries get removed from cdk.json. After running cdk synth my cdk.json looks like this:

"context": {
   "appName": "app1"
}

All values do persist in the cdk context though. Is this by design?

The problem im trying to solve is to have one repository, which can be deployed to a any environment by say doing - cdk deploy -c "ENV:QA".

Thanks!

Environment

• CDK CLI Version: 1.3.0 (build bba9914)
• Module Version:
• OS: Windows 10
• Language: TypeScript

Other information

[jgondron]

I think this should work, although this will break some of the cli options for managing context since cdk currently doesn’t really support this. In your case, I think you would need to do something like cdk deploy -c env="QA", then in code use that to look up the correct context object, ex:

const targetEnv = app.node.tryGetContext('env');
const env = app.node.tryGetContext(`ENV:${targetEnv}`);
const vpcId = env.vpcId;

IMO, it would be nice if there were an option similar to -c that would allow passing a json file to specify context for that specific deploy event. That way you could just have three separate files with similar structure, and not really have to do the extra gymnastics to process the target env inside of your code.

Ie, instead of the code above, you would be able to do something like cdk deploy --context-file env/qa.json where env/qa.json is something like:

{
  "appName": "app1",
  "vpcId": "vpc-222222222222"
}

then the code above is simplified to just:

const vpcId = app.node.tryGetContext('vpcId');

Doing this would also better support the use case of multiple devs working on similar parts of an application within the same account. A dev could have a personal file used to deploy their own dev env during iteration and exclude it from the repo.

[drissamri]

Is there a cleaner way to do this in meanwhile? It seems like a basic requirement for anything over POC’s/demos.

In our case we are deploying from Jenkins/GitLab and we have an AWS account (with different params) per environment so it should be easy to pass these in.