You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by "Jon Davies (JIRA)" <ji...@apache.org> on 2018/08/20 09:34:00 UTC
[jira] [Commented] (AIRFLOW-2920) Kubernetes pod operator:
namespace is a hard requirement
[ https://issues.apache.org/jira/browse/AIRFLOW-2920?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16585694#comment-16585694 ]
Jon Davies commented on AIRFLOW-2920:
-------------------------------------
Workaround appears to be to set
{code:java}
namespace=os.getenv('AIRFLOW_KUBE_NAMESPACE'),
{code}
...in the DAG code and then specify that variable in the scheduler configuration.
> Kubernetes pod operator: namespace is a hard requirement
> --------------------------------------------------------
>
> Key: AIRFLOW-2920
> URL: https://issues.apache.org/jira/browse/AIRFLOW-2920
> Project: Apache Airflow
> Issue Type: Bug
> Reporter: Jon Davies
> Priority: Major
>
> Hello,
> I'm using the Kubernetes pod operator for my DAGs, I install Airflow to its own namespace within my Kubernetes cluster (for example: "testing-airflow") and I would like pods spun up by that Airflow instance to live in that namespace.
> However, I have to hardcode the namespace into my DAG definition code and so I have to rebuild the Docker image for Airflow to be able to spin up a "production-airflow" namespace as the namespace is a hard requirement in the Python code - it'd be nice if the DAG could just default to its own namespace if none is defined.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)