You can add one or more node groups to your cluster if you do not have the right worker
node hardware (for example, incorrect GPU models for a new workload) in your existing cluster
configuration.
-
Generate the JSON skeleton to create the instance:
$ cdp ml add-instance-groups-ml-serving-app --generate-cli-skeleton > /tmp/add-instance-group.json
-
Edit the file to add the node group details:
{
"appCrn": "[***APPLICATION-CRN***]",
"instanceGroups": [
{
"instanceType": "g5.8xlarge",
"instanceCount": 0,
"rootVolume": {
"size": 1024
},
"autoscaling": {
"minInstances": 0,
"maxInstances": 4,
"enabled": true
}
},
{
"instanceType": "p3.8xlarge",
"instanceCount": 0,
"rootVolume": {
"size": 1024
},
"autoscaling": {
"minInstances": 0,
"maxInstances": 4,
"enabled": true
}
}
]
}
-
Use the JSON file created in the previous step to add the node group:
$ cdp ml add-instance-groups-ml-serving-app --cli-input-json file:///tmp/add-instance-group.json