gce_vm_create() R function from [googleComputeEngineR]

Creates an instance resource in the specified project using the data included in the request.


gce_vm_create(name, predefined_type = "f1-micro",
  image_project = "debian-cloud", image_family = "debian-8",
  cpus = NULL, memory = NULL, image = "", disk_source = NULL,
  network = "default", externalIP = NULL, canIpForward = NULL,
  description = NULL, metadata = NULL, scheduling = NULL,
  serviceAccounts = NULL, tags = NULL, minCpuPlatform = NULL,
  project = gce_get_global_project(), zone = gce_get_global_zone(),
  dry_run = FALSE, disk_size_gb = NULL, use_beta = FALSE,
  acceleratorCount = NULL, acceleratorType = "nvidia-tesla-p4")

Arguments

name: The name of the resource, provided by the client when initially creating the resource
predefined_type: A predefined machine type from gce_list_machinetype
image_project: Project ID of where the image lies
image_family: Name of the image family to search for
cpus: If not defining predefined_type, the number of CPUs
memory: If not defining predefined_type, amount of memory
image: Name of the image resource to return
disk_source: Specifies a valid URL to an existing Persistent Disk resource.
network: The name of the network interface
externalIP: An external IP you have previously reserved, leave NULL to have one assigned or "none" for no external access.
canIpForward: Allows this instance to send and receive packets with non-matching destination or source IPs
description: An optional description of this resource
metadata: A named list of metadata key/value pairs assigned to this instance
scheduling: Scheduling options for this instance, such as preemptible instances
serviceAccounts: A list of service accounts, with their specified scopes, authorized for this instance
tags: A list of tags to apply to this instance
minCpuPlatform: Specify a minimum CPU platform as per [https://cloud.google.com/compute/docs/instances/specify-min-cpu-platform](these Google docs)
project: Project ID for this request
zone: The name of the zone for this request
dry_run: whether to just create the request JSON
disk_size_gb: If not NULL, override default size of the boot disk (size in GB)
use_beta: If set to TRUE will use the beta version of the API. Should not be used for production purposes.
acceleratorCount: Number of GPUs to add to instance. If using this, you may want to instead use gce_vm_gpu which sets some defaults for GPU instances.
acceleratorType: Name of GPU to add, see gce_list_gpus

Returns

A zone operation, or if the name already exists the VM object from gce_get_instance

Details

Authentication scopes used by this function are:

cpus must be in multiples of 2 up to 32 memory must be in multiples of 256

One of image or image_family must be supplied

To create an instance you need to specify:

Name
Project [if not default]
Zone [if not default]
Machine type - either a predefined type or custom CPU and memory
Network - usually default, specifies open ports etc.
Image - a source image containing the operating system

You can add metadata to the server such as startup-script and shutdown-script. Details available here: https://cloud.google.com/compute/docs/storing-retrieving-metadata

If you want to not have an external IP then modify the instance afterwards

Preemptible VMS

You can set preemptible VMs by passing this in the scheduling arguments scheduling = list(preemptible = TRUE)

This creates a VM that may be shut down prematurely by Google - you will need to sort out how to save state if that happens in a shutdown script etc. However, these are much cheaper.

GPUs

Some defaults for launching GPU enabled VMs are available at gce_vm_gpu

You can add GPUs to your instance, but they must be present in the zone you have specified - use gce_list_gpus to see which are available. Refer to this link for a list of current GPUs per zone.

gce_vm_create function