[
https://issues.apache.org/jira/browse/AIRFLOW-3347?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16687005#comment-16687005
]
Chris Bandy commented on AIRFLOW-3347:
--------------------------------------
I was able to track this down. The following in kubernetes_executor was always
returning empty OrderedDict()
{code:python}
self.kube_secrets = configuration_dict.get('kubernetes_secrets', {})
{code}
The following line in AirflowConfigParser was not prepared for double
underscores within/beneath a configuration section. Limiting the split seemed
to do the trick.
{code:python}
diff --git a/airflow/configuration.py b/airflow/configuration.py
index 2e05fde0..4c923b80 100644
--- a/airflow/configuration.py
+++ b/airflow/configuration.py
@@ -358,7 +358,7 @@ class AirflowConfigParser(ConfigParser):
# add env vars and overwrite because they have priority
for ev in [ev for ev in os.environ if ev.startswith('AIRFLOW__')]:
try:
- _, section, key = ev.split('__')
+ _, section, key = ev.split('__', 2)
opt = self._get_env_var_option(section, key)
except ValueError:
opt = None
{code}
> Unable to configure Kubernetes secrets through environment
> ----------------------------------------------------------
>
> Key: AIRFLOW-3347
> URL: https://issues.apache.org/jira/browse/AIRFLOW-3347
> Project: Apache Airflow
> Issue Type: Bug
> Components: configuration, kubernetes
> Affects Versions: 1.10.0
> Reporter: Chris Bandy
> Priority: Major
>
> We configure Airflow through environment variables. While setting up the
> Kubernetes Executor, we wanted to pass the SQL Alchemy connection string to
> workers by including it the {{kubernetes_secrets}} section of config.
> Unfortunately, even with
> {{AIRFLOW_\_KUBERNETES_SECRETS_\_AIRFLOW_\_CORE_\_SQL_ALCHEMY_CONN}} set in
> the scheduler environment, the worker gets no environment secret environment
> variables.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
