第 14 章 Scaling Multicloud Object Gateway performance

The number of MultiCloud Object Gateway (MCG) endpoints scale automatically when the load on the MCG S3 service increases or decreases. OpenShift Data Foundation clusters are deployed with one active MCG endpoint. Each MCG endpoint pod is configured by default with 1 CPU and 2Gi memory request, with limits matching the request. When the CPU load on the endpoint crosses over an 80% usage threshold for a consistent period of time, a second endpoint is deployed lowering the load on the first endpoint. When the average CPU load on both endpoints falls below the 80% threshold for a consistent period of time, one of the endpoints is deleted. This feature improves performance and serviceability of the MCG.

You can scale the Horizontal Pod Autoscaler (HPA) for noobaa-endpoint using the following oc patch command, for example:

# oc patch -n openshift-storage storagecluster ocs-storagecluster \
    --type merge \
    --patch '{"spec": {"multiCloudGateway": {"endpoints": {"minCount": 3,"maxCount": 10}}}}'

The example above sets the minCount to 3 and the maxCount to `10.