durera
commented
1 day ago Removing cpu/memory requests and limits is what Best Effort QoS means.
- https://kubernetes.io/docs/concepts/workloads/pods/pod-qos/
- https://kubernetes.io/docs/tasks/configure-pod-container/quality-service-pod/#create-a-pod-that-gets-assigned-a-qos-class-of-besteffort
This is an advanced topic, and expectation is that the admin knows what they are doing/knows what the QoS they have chosen means.
The performance team recommends to put a caveat for users on that when choosing the Best Effort QoS, it may result in unbounded resource utilization.
We already provide a link to the documentation that explains what the QoS levels mean:
self.printDescription([
"The CLI supports two pod template profiles out of the box that allow you to reconfigure MAS for either a guaranteed or best effort QoS level",
"For more information about the Kubernetes quality of service (QoS) levels, see https://kubernetes.io/docs/concepts/workloads/pods/pod-qos/",
"You may also choose to use your own customized pod template definitions"
])
Choosing the "Best Effort QoS" in configuring pod templates in the MAS CLI results into the Suite CR having no cpu/memory requests/limits for all pods in the MAS core namespace. This is not recommended, as it causes coreapi pods to consume too much memory (>5GB each for a total of at least 15GB). Please refer to skills case TS017756933.
The performance team recommends to put a caveat for users on that when choosing the Best Effort QoS, it may result in unbounded resource utilization.