[ https://issues.apache.org/jira/browse/AIRFLOW-3819?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17005898#comment-17005898 ]
Lihan Li edited comment on AIRFLOW-3819 at 12/31/19 2:38 AM: ------------------------------------------------------------- Also just hit by this issue. I think this better resolved within Airflow, it is a very small addition; Also, Mesos has task_cpu within Airflow. [~aizhamal] was (Author: lihan): I think this better resolved within Airflow, it is a very small addition; Also, Mesos has task_cpu within Airflow. > k8s executor - Allow the configuration of a global default for pod resource > request/limits > ------------------------------------------------------------------------------------------- > > Key: AIRFLOW-3819 > URL: https://issues.apache.org/jira/browse/AIRFLOW-3819 > Project: Apache Airflow > Issue Type: Improvement > Components: executors > Affects Versions: 1.10.2 > Reporter: afusr > Priority: Minor > Labels: kubernetes > > Currently the kubernetes executor allows you to specify pod resources > requests and limits (cpu and memory). For example: > {noformat} > # Limit resources on this operator/task with node affinity & tolerations > three_task = PythonOperator( > task_id="three_task", python_callable=print_stuff, dag=dag, > executor_config={ > "KubernetesExecutor": {"request_memory": "128Mi", > "limit_memory": "128Mi", > "tolerations": tolerations, > "affinity": affinity}} > ) > {noformat} > These values are used by kubernetes when making scaling decisions. It would > be nice to be able to specify a global default for these values, to ensure > that each pod airflow creates has a value specified for these properties. > There is still the requirement to override these values on a dag by dag > basis. -- This message was sent by Atlassian Jira (v8.3.4#803005)