Chapter 15. Configuring Queue Proxy resources
The Queue Proxy is a sidecar container to each application container within a service. It improves managing Serverless workloads, ensuring efficient resource usage. You can configure the Queue Proxy.
15.1. Configuring Queue Proxy Resources for a Knative Service
Apart from configuring the Queue Proxy resource requests and limits globally in the deployment configmap, you can set them at the service level using the corresponding annotations targeting CPU, memory, and ephemeral storage resource types.
Prerequisites
- Red Hat OpenShift Pipelines must be installed on your cluster.
-
You have installed the OpenShift (
oc
) CLI. -
You have installed the Knative (
kn
) CLI.
Procedure
Modify the configmap of your service with resource requests and limits:
apiVersion: serving.knative.dev/v1 kind: Service metadata: name: example-service namespace: default spec: template: metadata: annotations: queue.sidecar.serving.knative.dev/cpu-resource-request: "1" queue.sidecar.serving.knative.dev/cpu-resource-limit: "2" queue.sidecar.serving.knative.dev/memory-resource-request: "1Gi" queue.sidecar.serving.knative.dev/memory-resource-limit: "2Gi" queue.sidecar.serving.knative.dev/ephemeral-storage-resource-request: "400Mi" queue.sidecar.serving.knative.dev/ephemeral-storage-resource-limit: "450Mi"
Alternatively, you can use the special
queue.sidecar.serving.knative.dev/resource-percentage
annotation, which calculates the Queue Proxy resources as a percentage of the application container. When CPU and memory resource requirements are calculated from the application container requirements and they are outside the boundaries below, the values are adjusted to fit within the boundaries. In this case, the following minimum and maximum boundaries are applied to the CPU and memory resource requirements:Table 15.1. Resource requirements boundaries Resource requirements Min Max CPU request
25m
100m
CPU limit
40m
500m
Memory request
50Mi
200Mi
Memory limit
200Mi
500Mi
NoteIf you simultaneously set a percentage annotation and a specific resource value using the corresponding resource annotation, then the latter takes precedence.
WarningThe
queue.sidecar.serving.knative.dev/resource-percentage
annotation is now deprecated and will be removed in a future version of OpenShift Serverless.