Kubernetes Executor -> is forbidden: exceeded quota #36356
-
Apache Airflow versionOther Airflow 2 version (please specify below) If "Other Airflow 2 version" selected, which one?2.7.1 What happened?I’m using the Kubernetes executor along with the Kubernetes pod operator. My DAG has many tasks running in parallel. However, when resources are not available for a task, I encounter an error. The error message indicates that the pod has exceeded its resource quota.
What you think should happen instead?I was under the impression that when using the Kubernetes executor, tasks would queue until sufficient resources become available, rather than failing. I’m wondering if I’ve overlooked some configuration settings. For most configurations, I’m using the default values provided by the Airflow Helm chart. Can anyone help me understand this? How to reproduceuse kubernetes and the airflow helmchart, set some resource quota for the namespace and try to execute a lot of tasks in parallel that would normally consume more resource than it is available Operating Systemubuntu Versions of Apache Airflow Providers2.7.1 DeploymentOfficial Apache Airflow Helm Chart Deployment detailsNo response Anything else?No response Are you willing to submit PR?
Code of Conduct
|
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 8 replies
-
Thanks for opening your first issue here! Be sure to follow the issue template! If you are willing to raise PR to address this issue please do so, no need to wait for approval. |
Beta Was this translation helpful? Give feedback.
-
hi @devscheffer , I believe this issue is related to your k8s cluster rather than to Airflow. You can check it with |
Beta Was this translation helpful? Give feedback.
-
Correct. If your task exceeds quota, things will fail. This is normal and expected. Converting to discussion if needed. |
Beta Was this translation helpful? Give feedback.
-
Hi @devscheffer , Did you find or propose any solution so far? This makes it exeptionally harder to run workloads via k8s exector and we had to heavily limit the number of pararell possible to tasks to mitigate this because it´s not clear from the outside how big the requirements of one task are. Thanks in advance. Best regards |
Beta Was this translation helpful? Give feedback.
Again - not sure about consequences, proposing it in PR is a good start though. Or if you do not know how starting a discussion specific for that proposal - possibly even starting discussion in devlist https://airflow.apache.org/community/ as KPO is one of the most used operator and posssibly there are other consequences of what you propose. I think there were also other discussions and issues about it and you can search through the issues and discussions for similar things.