[GitHub] [airflow] potiuk commented on issue #17489: Helm chart converts ' ' to '+' in passwords (and username)

[GitBox <gi...@apache.org>]

potiuk commented on issue #17489:
URL: https://github.com/apache/airflow/issues/17489#issuecomment-899311549


   > * **Whats the rationale for constructing and encoding the URL to be saved and putting it in the Secret?** This is as opposed to plopping the secret as provided by the user in the values.yaml and then building and encoding the URL inside the containers using some standard encoding & url construction strategy. Is it because this seemed like wasted work done on every pod reboot?
   
   Because Airflow takes URL as parameter and it's standard way for airflow (and airflow does it because this is the way how sqlalchemy does it) on one hand, or because it makes a unified way of passing not only the host/user/password/ but also extra parameters. If we want to construct the URL from secrets to Airflow, we would need some extra code to do that on the flight, rather than mapping the secret. In Airflow you simply "speak" URL when it comes to connections. You have to remember that Helm Chart and K8S are one of several ways Airflow can be deployed and in our choice we are airflow-centric not K8S centric. We feel that it's K8S that should bend to airflow ways, not the other way. You can see more at our talk from this year's summit: https://airflowsummit.org/sessions/2021/airflow-loves-kubernetes/
   
   > Do we think that a preconnect check should happen in the migration pod? I thought it was strange that it passed just fine and I didn't see connection errors until we were to the worker-pods this is why I initially assumed that some extra funky encoding was happening again behind the scenes and why using spaces raw seemed to work. I also thought it was strange that I did not see any of the other pods in a crashloop as well. Please correct me if I'm wrong about this.
   
   Of course it should. And maybe you discovered an issue there. Maybe you can track it down and provide a PR fixing that? That would be fantastic contribution back that would "pave the way" for others who have similar problem. Airflow is built by community and it's super helpful if users who get it for free are contributing back those kind of things.
   
   
   


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@airflow.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org