FloChehab commented on pull request #12010:
URL: https://github.com/apache/airflow/pull/12010#issuecomment-757358852


   > We run a set of integration tests for Kubernetes for each PR. To do it, we 
run a KIND cluster on Github Action, and in it, we deploy Airflow using the 
Helm Chart. After deployment, we perform a few simple operations, e.g. run DAG 
to check if everything works. After implementation, we perform a few simple 
operations, e.g. run DAG to check if everything works. To be able to test the 
Helm Chart with more confidence, I want to work on system tests that deploy 
infrastructure with Terraform and then install Airflow with Helm Chart. And 
then it does some basic operations.
   
   This feels quite ambitious and I personally thought the CI with a KIND 
cluster would be sufficient (for testing the integration with "external" 
services, you could start them separately in the same cluster) and relatively 
time & cost effective.
   
   To be honest, at the moment I am not sure I would give a hand on such a 
subject in the coming months ; I am trying to do bit less infrastructure 
related work these days :) (but I'll keep contributing on the helm chart).
   
   
   > Some changes would be useful in order for this change to be accepted. 
First of all, we should delete references to Google solutions, but the ability 
to configure your own annotations sounds interesting.
   > For example, if you want to use Workflow Identity then you need to 
annotate Service Account, so we don't want to support just one annotation, but 
we've added generic support for any annotation.
   > See:
   > #11131
   > #11387
   > 
   > Before this change was posted to the community, the change looked like 
this:
   > ![Screenshot 2021-01-09 at 19 02 
30](https://user-images.githubusercontent.com/12058428/104105300-6313e480-52ad-11eb-9212-6eede90d1e44.png)
   > It's not something we want to accept, so we had to rewrite it.
   
   Thanks for the info! I'll look into this in the coming weeks.
   
   > Additional facilities, if needed, can be deployed independently of Helm 
Chart.
   
   I was thinking about this, and I feel it's pretty annoying and harder to do 
/ maintain in a production setting; unless maybe if we go for some subchart 
mechanism (which I am not familiar with) or something else? I link when related 
things are bounded together; that's why I added for example the ability to 
create configmaps & secrets (for airflow) from within this chart: 
https://github.com/apache/airflow/pull/12164. 
   
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to