AWS CLI, using `--cli-input-json` in a pipeline

Solution 1:

Found a workaround for the time being with xargs that is quite clean:

cat ./mytask.json \
    | xargs -0 aws ecs register-task-definition --cli-input-json

It only adds xargs -0 and requires --cli-input-json to be the last argument

Solution 2:

I went digging... It looks like aws will read the indicated file twice, using the second dataset for it's operation. Of course, in a pipeline, the second read() will get nothing.

I've added a pipe:// prefix/schema (commit) for use in this situation which will cache the value... I've also made a pull request.

Solution 3:

I was able to pass the JSON as a variable on --cli-input-json and inject bash variables too.

So your example should be:

aws ecs register-task-definition --cli-input-json "$(cat < ./mytask.json)"

In my scenario at hand I have:

1. Either pass a JSON as you would like

   aws autoscaling start-instance-refresh --cli-input-json "$(cat < ./options.json )"

2. Or as HEREDOC in order to pass other variables too. (Not requested but may be useful)

asg_nominal_name=AutoScalingGroupName

aws autoscaling start-instance-refresh --cli-input-json "$(cat <<JSON
{ 
    "AutoScalingGroupName": "${asg_nominal_name}",
    "Preferences": {
        "InstanceWarmup": 300,
        "MinHealthyPercentage": 100
    }
}
JSON
)"

I am not so familiar to explain the why in bash terms. Feel free.