[SOLVED] Pod shell - modifying the already declared ENV variables

Issue

Context/why am I looking for that strange way:

the pod (pods) where I'm trying to inject the envs are spawned by a job and there are hundreds of them. There is also a configmap in k8s which holds all the envs. I'm doing this to avoid manually declaring hundreds of pods with different env vars

I have a pod in k8s with defined env vars in the manifest:

spec:
  containers:
    name: something
    command: ["/bin/bash"]
    args: ["-c", "sleep 60 && source ~/.bashrc && env | grep FOO"]
    env:
    - name: FOO
      value: bar
    - name: FOO2
      value: bar2

I also run another pod with kubectl and rbac set that executes:

kubectl  exec -i FIRST_POD_NAME -- bash -c "echo 'FOO=updated_bar' >> ~/.bashrc"

first pods use (sleep is to ensure the other pods has time to update the values of env vars):

sleep 60 && source ~/.bashrc && env | grep FOO

as command and it logs:

FOO=bar

FOO2=bar2

but when I manually enter the pod (bash) and run:

env | grep FOO

I see FOO=updated_bar

What am I missing here and how to update these envs so the pod uses them in start cmd?

Solution

As I was unable to find a reason for this, I've made a workaround so if anyone needs to do something similar:

helping pod is just appending the values to tmp files in other pods:

kubectl  exec -i FIRST_POD_NAME -- bash -c "echo 'FOO=updated_bar' > /tmp/foo

and on the startup I read the value of this fire and use it as ENV:

# some sleep is needed to let the helping pod created this file
sleep n && export FOO=$(cat /tmp/foo)

this is a poor, but working solution to import envs declared in configmap to hundrets of pods

Answered By - wiktor

Answer Checked By - Pedro (WPSolving Volunteer)