[
https://issues.apache.org/jira/browse/AIRFLOW-2917?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16877788#comment-16877788
]
Dimitris Theodorou commented on AIRFLOW-2917:
---------------------------------------------
This issue is resolved in a bogus way. It assumes that if a configmap is set,
then that configmap contains the `AIRFLOW__CORE____SQL_ALCHEMY_CONN` and does
not pick up the connection string from the executor's configuration. So when a
configmap is specified without a connection string, the workers end up trying
to use the bogus default sqlite connection.
The problem is that it is almost always a bad practice to store secrets in a
configmap in the context of Kubernetes secret management. The connection string
itself is a secret and should be passed as a secret, external to the configmap.
So I consider the use-case of supporting passing the connection string from the
configmap of much less importance than supporting the use-case of passing it
from the executor.
I think this use case should still be supported, but the configmap should be
explicitly checked that it contains a connection string, and if not fall back
to using the executor's
> Set AIRFLOW__CORE__SQL_ALCHEMY_CONN only when needed for k8s executor
> ---------------------------------------------------------------------
>
> Key: AIRFLOW-2917
> URL: https://issues.apache.org/jira/browse/AIRFLOW-2917
> Project: Apache Airflow
> Issue Type: Improvement
> Components: executors
> Affects Versions: 1.10.0
> Reporter: John Cheng
> Assignee: John Cheng
> Priority: Minor
> Fix For: 1.10.2
>
>
> In Kubernetes executor, `AIRFLOW__CORE__SQL_ALCHEMY_CONN` is set as an
> environment variable even when it is specified in configmap or secrets.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
