yehoshuadimarsky commented on issue #20408: URL: https://github.com/apache/airflow/issues/20408#issuecomment-998437846
> > I see. The docs are confusing in that regard then, I thought the `remote_log_conn_id` is used for both reading and writing - for all operations for remote logging. > > I heartily invite you to make a PR that will make it clearer there. Airflow is created by > 1800 contributors - most of them contributing code and docs voluntarily and this is an easy way to become one of the contributors and "give back" for the software you get for free. This is super easy - you just use "Suggest a change on this page" button at the bottom-right of the doc page and you will get a possibility to submit changes to the dodcumentation using GitHub UI as a new PR. It's even easier to make a PR than to create an issue about it in fact. And you as the user are one of the best to know what and how to describe it in the way that other users will find useful. Thank you! I will definitely try to submit a PR to change the docs, thank you for making it easy and inviting. > > > So maybe I can broaden my question then - why are there 2 different connection IDs, `remote_log_conn_id` and `google_key_path`? Maybe the GCP provider package should change the remote logging to use `remote_log_conn_id` for both read and write? > > I belive (I have not implemented it) this is because of security. The conn_id is there and (currently) available to anyone who has access to the DB (including the webserver to read the logs). However writing should be exclusively available to the tasks "workload". Write access to the bucket is generally more "precious" and for example workload identity and using metadata server in GCP gives much better security, because the access to write is only granted to that particular machine executing the workload and only for the time of executing the task. Even if such "authtentication" leaks, it is useless outside of this single task execution. On the other hand the "conn_id" details are static and leaking it could give anyone unlimited and uncontrolled access to write any data to the bucket. Which could lead to a number of security issues (such as log injection vulnerabilities etc). This makes sense, thank you. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
