Run `gcloud auth appplication-default login
to create your credentials. (Alternatively, you may use a service account, see Using a Service Account section).
Add the public SSH key you're going to use to GCE Metadata in Compute ->
Compute Engine -> Metadata section of the console, SSH Keys tab. (Read
the SSH Support
readme section for more information.)
Using a Service Account
The `appplication-default login
method is intended to be used for developing code on a local environment - this is typically Vagrant's use-case
as well. However, if this is not your use-case, you will want to use a credential not tied to your local environment:
a service account.
To use a service account:
While still in the API & Services, go to
Credentials subsection,
and click Create credentials -> Service account.
Create a Service Account with any name (f.e. vagrant) and grant it
a Compute Admin role.
Open the new service account page and click on the Keys tab.
Click Add key -> Create new key, choose JSON. Download the JSON private key
and save this file in a secure and reliable location.
Then include the private key in your Vagrantfile's provider block as a google_json_key_location
attribute:
Vagrant.configure("2")do |config|
# ... other stuffconfig.vm.provider:googledo |google|
google.google_project_id="YOUR_GOOGLE_CLOUD_PROJECT_ID"google.google_json_key_location="/path/to/your/private-key.json"endend
Vagrant Setup
Install as a Vagrant plugin:
vagrant plugin install vagrant-google
Usage
Make a Vagrantfile that looks like the following, filling in
your information where necessary:
This will start the latest version of Ubuntu 20.04 LTS instance in the
us-central1-f zone, with an n1-standard-1 machine, and the "default"
network within your project. And assuming your SSH information (see below) was
filled in properly within your Vagrantfile, SSH and provisioning will work as
well.
Note that normally a lot of this boilerplate is encoded within the box file,
but the box file used for the quick start, the "google" box, has no
preconfigured defaults.
SSH Support
In order for SSH to work properly to the GCE VM, you will first need to add
your public key to the GCE metadata service for the desired VM user account.
When a VM first boots, a Google-provided daemon is responsible for talking to
the internal GCE metadata service and creates local user accounts and their
respective ~/.ssh/authorized_keys entries. Most new GCE users will use the
Cloud SDKgcloud compute utility when first
getting started with GCE. This utility has built in support for creating SSH
key pairs, and uploading the public key to the GCE metadata service. By
default, gcloud compute creates a key pair named
~/.ssh/google_compute_engine[.pub].
Note that you can use the more standard ~/.ssh/id_rsa[.pub] files, but you
will need to manually add your public key to the GCE metadata service so your
VMs will pick up the key. Note that the public key is typically
prefixed with the username, so that the daemon on the VM adds the public key
to the correct user account.
Additionally, you will probably need to add the key and username to override
settings in your Vagrantfile like so:
config.vm.provider:googledo |google,override|
#...google provider settings are skipped...override.ssh.username="testuser"override.ssh.private_key_path="~/.ssh/id_rsa"#...google provider settings are skipped...end
See the links below for more help with SSH and GCE VMs.
Every provider in Vagrant must introduce a custom box format. This provider
introduces google boxes. You can view an example box in
example_boxes/.
That directory also contains instructions on how to build a box.
The box format is basically just the required metadata.json file along with
a Vagrantfile that does default settings for the provider-specific
configuration for this provider.
Configuration
This provider exposes quite a few provider-specific configuration options:
google_json_key_location - The location of the JSON private key file matching your
Service Account.
(Can also be configured with GOOGLE_JSON_KEY_LOCATION environment variable.)
google_project_id - The Project ID for your Google Cloud Platform account.
(Can also be configured with GOOGLE_PROJECT_ID environment variable.)
image - The image name to use when booting your instance.
image_family - Specify an "image family" to pull the latest image from. For example: centos-7
will pull the most recent CentOS 7 image. For more info, refer to
Google Image documentation.
image_project_id - The ID of the GCP project to search for the image or image_family.
For example: centos-cloud for Centos 7/8/Stream image families.
instance_group - Unmanaged instance group to add the machine to. If one
doesn't exist it will be created.
instance_ready_timeout - The number of seconds to wait for the instance
to become "ready" in GCE. Defaults to 20 seconds.
machine_type - The machine type to use. The default is "n1-standard-1".
disk_size - The disk size in GB. The default is 10.
disk_name - The disk name to use. If the disk exists, it will be reused, otherwise created.
disk_type - Whether to use Standard disk or SSD disk. Use either pd-ssd or pd-standard.
autodelete_disk - Boolean whether to delete the disk when the instance is deleted or not. Default is true.
metadata - Custom key/value pairs of metadata to add to the instance.
name - The name of your instance. The default is "i-yyyymmddhh-randomsd",
e.g. 10/08/2015 13:15:15 is "i-2015081013-15637fda".
network - The name of the network to use for the instance. Default is
"default".
network_project_id - The ID of the GCP project for the network and subnetwork to use for the instance. Default is google_project_id.
subnetwork - The name of the subnetwork to use for the instance.
tags - An array of tags to apply to this instance.
labels - Custom key/value pairs of labels to add to the instance.
zone - The zone name where the instance will be created.
can_ip_forward - Boolean whether to enable IP Forwarding.
external_ip - The external IP address to use (supports names). Set to false to not assign an external address.
network_ip - The internal IP address to use. Default is to use next available address.
use_private_ip - Boolean whether to use private IP for SSH/provisioning. Default is false.
preemptible - Boolean whether to enable preemptibility. Default is false.
auto_restart - Boolean whether to enable auto_restart. Default is true.
on_host_maintenance - What to do on host maintenance. Can be set to MIGRATE or TERMINATE Default is MIGRATE.
scopes or service_accounts - An array of OAuth2 account scopes for
services that the instance will have access to. Those can be both full API
scopes, just endpoint aliases (the part after ...auth/), and gcloud
utility aliases, for example:
['storage-full', 'bigquery', 'https://www.googleapis.com/auth/compute'].
service_account - The IAM service account email to use for the instance.
additional_disks - An array of additional disk configurations. disk_size is default to 10GB;
disk_name is default to name + "-additional-disk-#{index}"; disk_type is default to pd-standard;
autodelete_disk is default to true. Here is an example of configuration.
accelerators - An array of accelerator configurations. type is the
accelerator type (e.g. nvidia-tesla-k80); count is the number of
accelerators and defaults to 1. Note that only TERMINATE is supported for
on_host_maintenance; this should be set explicitly, since the default is
MIGRATE.
enable_secure_boot - For Shielded VM, whether to enable Secure Boot.
enable_vtpm - For Shielded VM, whether to enable vTPM.
enable_integrity_monitoring - For Shielded VM, whether to enable Integrity monitoring.
These can be set like typical provider-specific configuration:
Vagrant.configure("2")do |config|
# ... other stuffconfig.vm.provider:googledo |google|
google.google_project_id="YOUR_GOOGLE_CLOUD_PROJECT_ID"google.google_json_key_location="/path/to/your/private-key.json"endend
In addition to the above top-level configs, you can use the zone_config
method to specify zone-specific overrides within your Vagrantfile. Note
that the top-level zone config must always be specified to choose which
zone you want to actually use, however. This looks like this:
Vagrant.configure("2")do |config|
config.vm.box="google/gce"config.vm.provider:googledo |google|
google.google_project_id="YOUR_GOOGLE_CLOUD_PROJECT_ID"google.google_json_key_location="/path/to/your/private-key.json"# Make sure to set this to trigger the zone_configgoogle.zone="us-central1-f"google.zone_config"us-central1-f"do |zone1f|
zone1f.name="testing-vagrant"zone1f.image="debian-9-stretch-v20211105"zone1f.machine_type="n1-standard-4"zone1f.zone="us-central1-f"zone1f.metadata={'custom'=>'metadata','testing'=>'foobarbaz'}zone1f.scopes=['bigquery','monitoring','https://www.googleapis.com/auth/compute']zone1f.tags=['web','app1']endendend
The zone-specific configurations will override the top-level configurations
when that zone is used. They otherwise inherit the top-level configurations,
as you would expect.
Networking features in the form of config.vm.network are not supported
with vagrant-google, currently. If any of these are specified, Vagrant will
emit a warning, but will otherwise boot the GCE machine.
Synced Folders
Since plugin version 2.0, this is implemented via built-in SyncedFolders action.
See Vagrant's rsync action
documentation for more info.
Automatic shutdown
To save money you may want to ensure you don't forget to shut down your instances
when you stop using them.
A very basic solution for this is to use Vagrant's provisioning feature to plan
automatic shutdown of the vm after given time after each vagrant up:
# Plan automatic shutdown of machine to prevent unwanted costsconfig.vm.provision"auto-shutdown",type: "shell",run: "always",inline: "shutdown -P +480"# = 60 minutes * 8 hours
Print external IP
You may want to know your machine's external IP f.e. to put it in your Ansible inventory
or open the app you deploy in it in your browser.
To automate printing it IP you can also use the Vagrant's provisioning feature:
To work on the vagrant-google plugin, clone this repository, and use
Bundler to get the dependencies:
$ bundle
Once you have the dependencies, verify the unit tests pass with rake:
$ bundle exec rake
If those pass, you're ready to start developing the plugin. You can test
the plugin without installing it into your Vagrant environment by just
creating a Vagrantfile in the top level of this directory (it is ignored by
git), and use bundler to execute Vagrant:
$ bundle exec vagrant up --provider=google
Acceptance testing
Work-in-progress: Acceptance tests are based on vagrant-spec library which
is currently under active development so they may occasionally break.
Before you start acceptance tests, you'll need to set the authentication
shell variables accordingly:
请发表评论