You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
While Kubernetes itself isn't one of the biggest cost drivers, most of the fees it causes are reflected in the Compute Engine charges and the configuration of Kubernetes can be a significant driver of that cost. We identified that the data-infra-apps cluster (where users' JupyterHub environments run) is configured to be multi-zone.
We recommend scale the non-airflow cluster down to 1 zone so there is a single pool of nodes that everyone shares, and we could potentially cut the min number of nodes in half while also tuning the autoscaling to be more responsive.
The text was updated successfully, but these errors were encountered:
Chore
Review Kubernetes configuration.
While Kubernetes itself isn't one of the biggest cost drivers, most of the fees it causes are reflected in the Compute Engine charges and the configuration of Kubernetes can be a significant driver of that cost. We identified that the data-infra-apps cluster (where users' JupyterHub environments run) is configured to be multi-zone.
We recommend scale the non-airflow cluster down to 1 zone so there is a single pool of nodes that everyone shares, and we could potentially cut the min number of nodes in half while also tuning the autoscaling to be more responsive.
The text was updated successfully, but these errors were encountered: