Hi folks! I have a StepDecorator where I want to add some stuff to the environment wherever the task is run (which will usually be using

@kubernetes

in my case). I cribbed from what the

@environment

decorator does, and did something like this:

def runtime_step_cli(self, cli_args, *_args):
        if self.attributes["some_attribute"]:
            cli_args.env.update({"SOME_ENV_VAR": "SOME_VALUE"})

Then I have a

task_pre_step

that uses that env var. However, I'm not seeing

SOME_ENV_VAR

set in

os.environ

within the task_pre_step when running in a kubernetes pod. I have confirmed that it is present during

step_init

while the task is being started up on the local machine, and the values I'm setting using the

environment

decorator are making it to the pod, but not the ones from my extension. Any ideas what I'm missing here?

I'm not familiar with exactly how this works, but I have seen that the

@environment

decorator gets special treatment in various places. When we want a custom decorator to set environment variables, we typically have it modify the environment decorator rather than set the variables directly.

hey — your env won’t be pushed this way (the cli_args are only used to launch locally so not passed down).

Sorry for the radio silence, thanks @quiet-afternoon-68940 and @dry-beach-38304! Yep, I think this is what's going on! Gonna try that solution and report back.

cool. lmk

This is how I add env vars to every step: (what you're asking but similar ballpack)

def flow_init(self, flow, graph, environment, flow_datastore, metadata, logger, echo, options):

env_config = {

"NAME": "VALUE"

}

my_decorators = [

f"environment:vars={json.dumps(env_config)}",

]

_attach_decorators(flow, my_decorators)