mirror of
https://github.com/hanzo-ml/kuberay.git
synced 2026-10-30 19:48:47 +00:00
Google Cloud Storage (GCS)
This module is the writer and reader for GCS storage.
It is required for the GKE Cluster running Ray to have workload identity (WI), to setup WI, please follow: How-to: Workload Identity
To use it with the History Server, set --runtime-class-name=gcs.
apiVersion: apps/v1
kind: Deployment
...
spec:
...
template:
...
spec:
serviceAccountName: historyserver
containers:
- name: historyserver
env:
- name: GCS_BUCKET
value: "<GCS_BUCKET_NAME>"
image: ray-historyserver:v0.1.0
imagePullPolicy: Always
command:
- historyserver
- --runtime-class-name=gcs
- --ray-root-dir=log
ports:
- containerPort: 8080
resources:
limits:
cpu: "500m"
RayCluster will also have the following under both the worker and head collector spec
- name: collector
image: ray-collector:v0.1.0
imagePullPolicy: Always
env:
- name: GCS_BUCKET
value: "<GCS_BUCKET_NAME>"
command:
- collector
- --role=Head
- --runtime-class-name=gcs
- --ray-cluster-name=raycluster-historyserver
- --ray-root-dir=log
- --events-port=8084