CODEWITHSUNDEEP
google-cloud-platformserverlessgoogle-cloud-run
Adrian
Adrian

Reputation: 2113

Cloud Run Memory Restriction

How do I get a Cloud Run container that uses more than 2 GB run on custom GKE cluster?

Since Cloud Run uses knative, I wonder if it is possible to tweak deployment describtor with more allocated/allowed memory limit to run it on GKE.

apiVersion: serving.knative.dev/v1alpha1
kind: Revision
metadata:
 ...

Upvotes: 3

Views: 2497

Answers (2)

John Hanley
John Hanley

Reputation: 81336

How do I get a Cloud Run container that uses more than 2 GB run on custom GKE cluster?

The maximum memory that you can allocate to a container in Cloud Run Managed is 2 GB.

[UPDATE 2]

Cloud Run now supports 32 GB, but the larger memory features are in preview. This document provides details on memory limits.

Set and update memory limits

[UPDATE]

For Cloud Run on Kubernetes, you can request more memory:

gcloud beta run deploy --image gcr.io/cloudrun/hello --memory=4G --cluster ha-cluster-1

Since Cloud Run uses knative, I wonder if it is possible to tweak deployment describtor with more allocated/allowed memory limit to run it on GKE.

Cloud Run Managed does not run in Knative, it runs in gVisor. I wrote an article that describes the Cloud Run infrastructure plus the Knative API that both Cloud Run Managed and Cloud Run on Kubernetes expose here. However, even with direct access to the Cloud Run Knative API, you cannot get around the imposed service limits.

The purpose of Cloud Run is to simplify deployments by abstracting away implementation details of the underlying infrastructure. There are cases where you should deploy directly to Kubernetes.

Upvotes: 6

ahmet alp balkan
ahmet alp balkan

Reputation: 45196

As you said, when you are using Cloud Run on GKE, you're using Knative serving stack, which allows you to configure custom CPU and memory requests/limits just like Kubernetes: https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/

For example, to get 4 GiB memory, you must add a field like:

resources:
  limits:
    memory: "4GiB"

to your container: field in Knative Service spec.

In Knative 0.6 (serving v1beta1) this field's path would be Service.spec.template.spec.containers[*].resources field, in earlier versions (serving v1alpha1) something like Service.spec.runLatest.revisionTemplate.spec.container.resources.

Upvotes: 1

Related Questions