consul/website/content/docs/install/cloud-auto-join.mdx

---
layout: docs
page_title: Auto-join a Cloud Provider
description: >-
  Auto-join enables agents to automatically join other agents running in the cloud. To configure auto-join, specify agent addresses with compute node metadata for the cloud provider instead of an IP address. Use the CLI or an agent configuration file to configure cloud auto-join.
---

# Cloud Auto-join

As of Consul 0.9.1, `retry-join` accepts a unified interface using the
[go-discover](https://github.com/hashicorp/go-discover) library for
automatically joining a Consul datacenter using cloud metadata. To use `retry-join` with a
supported cloud provider, specify the configuration on the command line or
configuration file as a `key=value key=value ...` string.

In Consul 0.9.1-0.9.3 the values need to be URL encoded but for most
practical purposes you need to replace spaces with `+` signs.

As of Consul 1.0 the values are taken literally and must not be URL
encoded. If the values contain spaces, equals, backslashes or double quotes then
they need to be double quoted and the usual escaping rules apply.

```shell-session
$ consul agent -retry-join 'provider=my-cloud config=val config2="some other val" ...'
```

or via a configuration file:

```json
{
  "retry_join": ["provider=my-cloud config=val config2=\"some other val\" ..."]
}
```

## Auto-join with Network Segments <EnterpriseAlert inline />

In order to use cloud auto-join with [Network Segments](/docs/enterprise/network-segments),
you must reconfigure the Consul agent's Serf LAN port to match that of the
segment you wish to join.

For example, given the following segment configuration on the server agents:

<CodeBlockConfig filename="server-config.hcl">

```hcl
segments = [
  {
    name = "alpha"
    bind = "{{GetPrivateIP}}"
    advertise = "{{GetPrivateIP}}"
    port = 8303
  },
  {
    name = "beta"
    bind = "{{GetPrivateIP}}"
    advertise = "{{GetPrivateIP}}"
    port = 8304
  }
]
```

</CodeBlockConfig>

A Consul client agent wishing to join the "alpha" segment would need to be configured
to use port `8303` as its Serf LAN port prior to attempting to join the cluster.

<Tabs>
<Tab heading="Agent configuration">

The following example configuration overrides the default Serf LAN port using the
[`ports.serf_lan`](/docs/agent/config/config-files#serf_lan_port) configuration option.

<CodeBlockConfig filename="client-config.hcl">

```hcl
ports {
  serf_lan = 8303
}
```

</CodeBlockConfig>

</Tab>
<Tab heading="Command-line flag">

The following example overrides the default Serf LAN port using the
[`-serf-lan-port`](/docs/agent/config/cli-flags#_serf_lan_port) command line flag.

```shell
$ consul agent -serf-lan-port=8303 -retry-join "provider=..."
```

</Tab>
</Tabs>

## Provider-specific configurations

The cloud provider-specific configurations are detailed below. This can be
combined with static IP or DNS addresses or even multiple configurations
for different providers.

In order to use discovery behind a proxy, you will need to set
`HTTP_PROXY`, `HTTPS_PROXY` and `NO_PROXY` environment variables per
[Golang `net/http` library](https://golang.org/pkg/net/http/#ProxyFromEnvironment).

The following sections give the options specific to each supported cloud
provider.

### Amazon EC2 and ECS

This returns the first private IP address of all servers in the given
region which have the given `tag_key` and `tag_value`.

```shell-session
$ consul agent -retry-join "provider=aws tag_key=... tag_value=..."
```

```json
{
  "retry_join": ["provider=aws tag_key=... tag_value=..."]
}
```

- `provider` (required) - the name of the provider ("aws" in this case).
- `tag_key` (required) - the key of the tag to auto-join on.
- `tag_value` (required) - the value of the tag to auto-join on.
- `region` (optional) - the AWS region to authenticate in.
- `addr_type` (optional) - the type of address to discover: `private_v4`, `public_v4`, `public_v6`. Default is `private_v4`. (>= 1.0)
- `access_key_id` (optional) - the AWS access key for authentication (see below for more information about authenticating).
- `secret_access_key` (optional) - the AWS secret access key for authentication (see below for more information about authenticating).
- `service` (optional) - String value that specifies which AWS service to filter. You can specify either `ec2` or `ecs`. Default is `ec2`.
- `ecs_cluster` (optional) - String value that limits searches to a specific AWS ECS cluster name or full ARN. By default, Consul searches all clusters with the specified tag values.
- `ecs_family` (optional) - String value limits searches to a AWS ECS task definition family. By default, Consul searches all task definition families with the specified tags.
- `endpoint` (optional) - String value that specifies the endpoint URL of the AWS service to use. If not set, the AWS client sets the value, which defaults to the public DNS name for the service in the specified region.

#### Authentication & Precedence

- Static credentials `access_key_id=... secret_access_key=...`
- Environment variables (`AWS_ACCESS_KEY_ID` and `AWS_SECRET_ACCESS_KEY`)
- Shared credentials file (`~/.aws/credentials` or the path specified by `AWS_SHARED_CREDENTIALS_FILE`)
- ECS task role metadata (container-specific).
- EC2 instance role metadata.

The only IAM permission required for discovering EC2 consul-servers is `ec2:DescribeInstances`.
We recommend that you make a dedicated key used only to auto-join the datacenter.
If the region is omitted it will be discovered through the local instance's [EC2 metadata
endpoint](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/instance-identity-documents.html).

The AWS ECS task role associated with the service attempting to discover the `consul-server` must have the following IAM permissions:
- `ecs:ListClusters` (only used when `ecs_cluster` is not provided)
- `ecs:ListServices` (only used when `ecs_cluster` is not provided)
- `ecs:DescribeServices` (only used when `ecs_cluster` is not provided)
- `ecs:ListTasks`
- `ecs:DescribeTasks`
If the region is omitted from the configuration, Consul obtains it from the local instance's [ECS V4 metadata endpoint](https://docs.aws.amazon.com/AmazonECS/latest/developerguide/task-metadata-endpoint-v4.html).

### Microsoft Azure

This returns the first private IP address of all servers in the given region
which have the given `tag_key` and `tag_value` applied to their virtual NIC in the tenant and subscription, or in
the given `resource_group` of a `vm_scale_set` for Virtual Machine Scale Sets.

```shell-session
$ consul agent -retry-join "provider=azure tag_key=... tag_value=... tenant_id=... client_id=... subscription_id=... secret_access_key=..."
```

```json
{
  "retry_join": [
    "provider=azure tag_key=... tag_value=... tenant_id=... client_id=... subscription_id=... secret_access_key=..."
  ]
}
```

- `provider` (required) - the name of the provider ("azure" in this case).
- `tenant_id` (required) - the tenant to join machines in.
- `client_id` (required) - the client to authenticate with.
- `secret_access_key` (required) - the secret client key. **NOTE** This value often may have an equals sign in its value, especially if generated from the Azure Portal, so is important to wrap in single quotes eg. `secret_access_key='fpOfcHQJAQBczjAxiVpeyLmX1M0M0KPBST+GU2GvEN4='`

Variables can also be provided by environmental variables:

- `ARM_SUBSCRIPTION_ID` for subscription
- `ARM_TENANT_ID` for tenant
- `ARM_CLIENT_ID` for client
- `ARM_CLIENT_SECRET` for secret access key

Use these configuration parameters when using tags:

- `tag_key` - the name of the tag to auto-join on.
- `tag_value` - the value of the tag to auto-join on.

Use these configuration parameters (instead of `tag_name` and `tag_value`) when using Virtual Machine Scale Sets (Consul 1.0.3 and later):

- `resource_group` - the name of the resource group to filter on.
- `vm_scale_set` - the name of the virtual machine scale set to filter on.

When using tags the only permission needed is `Microsoft.Network/networkInterfaces/read`. The scope for this permission is the Resource Group
that contains the virtual NICs for the Virtual Machines.

When using Virtual Machine Scale Sets the only role action needed is `Microsoft.Compute/virtualMachineScaleSets/*/read`.

~> **Note:** If the Consul datacenter is hosted on Azure, Consul can use Managed Service Identities (MSI) to access Azure instead of an environment
variable, shared client id and secret. MSI must be enabled on the VMs or Virtual Machine Scale Sets hosting Consul. It is the preferred configuration
since MSI prevents your Azure credentials from being stored in Consul configuration. This feature is supported in Consul 1.7 and above. When using
MSI, the `tag_key`, `tag_value` and `subscription_id` need to be supplied for Virtual machines.
Be aware that the amount of time that Azure takes for the VMs to detect the MSI permissions can be between a minute to an hour.

### Google Compute Engine

This returns the first private IP address of all servers in the given
project which have the given `tag_value`.

```shell-session
$ consul agent -retry-join "provider=gce project_name=... tag_value=..."
```

```json
{
  "retry_join": ["provider=gce project_name=... tag_value=..."]
}
```

- `provider` (required) - the name of the provider ("gce" in this case).
- `tag_value` (required) - the value of the tag to auto-join on.
- `project_name` (optional) - the name of the project to auto-join on. Discovered if not set.
- `zone_pattern` (optional) - the list of zones can be restricted through an RE2 compatible regular expression. If omitted, servers in all zones are returned.
- `credentials_file` (optional) - the credentials file for authentication. Note, if you set `-config-dir` do not store the credentials.json file in the configuration directory as it will be parsed as a config file and Consul will fail to start. See below for more information.

#### Authentication & Precedence

Discovery requires a [GCE Service
Account](https://cloud.google.com/compute/docs/access/service-accounts).
Credentials are searched using the following paths, in order of precedence.

- Use credentials from `credentials_file`, if provided.
- Use JSON file from `GOOGLE_APPLICATION_CREDENTIALS` environment variable.
- Use JSON file in a location known to the gcloud command-line tool.
  - On Windows, this is `%APPDATA%/gcloud/application_default_credentials.json`.
  - On other systems, `$HOME/.config/gcloud/application_default_credentials.json`.
- On Google Compute Engine, use credentials from the metadata
  server. In this final case any provided scopes are ignored.

### IBM SoftLayer

This returns the first private IP address of all servers for the given
datacenter with the given `tag_value`.

```shell-session
$ consul agent -retry-join "provider=softlayer datacenter=... tag_value=... username=... api_key=..."
```

```json
{
  "retry_join": [
    "provider=softlayer datacenter=... tag_value=... username=... api_key=..."
  ]
}
```

- `provider` (required) - the name of the provider ("softlayer" in this case).
- `datacenter ((#sl_datacenter)) (required) - the name of the datacenter to auto-join in.
- `tag_value` (required) - the value of the tag to auto-join on.
- `username` (required) - the username to use for auth.
- `api_key` (required) - the api key to use for auth.

### Aliyun (Alibaba Cloud)

This returns the first private IP address of all servers for the given
`region` with the given `tag_key` and `tag_value`.

```shell-session
$ consul agent -retry-join "provider=aliyun region=... tag_key=consul tag_value=... access_key_id=... access_key_secret=..."
```

```json
{
  "retry_join": [
    "provider=aliyun region=... tag_key=consul tag_value=... access_key_id=... access_key_secret=..."
  ]
}
```

- `provider` (required) - the name of the provider ("aliyun" in this case).
- `region` (required) - the name of the region.
- `tag_key` (required) - the key of the tag to auto-join on.
- `tag_value` (required) - the value of the tag to auto-join on.
- `access_key_id` (required) -the access key to use for auth.
- `access_key_secret` (required) - the secret key to use for auth.

The required RAM permission is `ecs:DescribeInstances`.
It is recommended you make a dedicated key used to auto-join.

### Digital Ocean

This returns the first private IP address of all servers for the given
`region` with the given `tag_name`.

```shell-session
$ consul agent -retry-join "provider=digitalocean region=... tag_name=... api_token=..."
```

```json
{
  "retry_join": ["provider=digitalocean region=... tag_name=... api_token=..."]
}
```

- `provider` (required) - the name of the provider ("digitalocean" in this case).
- `region` (required) - the name of the region.
- `tag_name` (required) - the value of the tag to auto-join on.
- `api_token` (required) -the token to use for auth.

### Openstack

This returns the first private IP address of all servers for the given
`region` with the given `tag_key` and `tag_value`.

```shell-session
$ consul agent -retry-join "provider=os tag_key=consul tag_value=server user_name=... password=... auth_url=..."
```

```json
{
  "retry_join": [
    "provider=os tag_key=consul tag_value=server user_name=... password=... auth_url=..."
  ]
}
```

- `provider` (required) - the name of the provider ("os" in this case).
- `tag_key` (required) - the key of the tag to auto-join on.
- `tag_value` (required) - the value of the tag to auto-join on.
- `domain_name` (optional) - the name of the domain.
- `domain_id` (optional) - the id of the domain.
- `project_id` (optional) - the id of the project (tenant id).
- `region` (optional) - the name of the region.
- `user_name` (optional) - the username to use for auth.
- `password` (optional) - the password to use for auth.
- `token` (optional) - the token to use for auth.
- `auth_url` (optional) - the identity endpoint to use for auth.
- `insecure` (optional) - indicates whether the API certificate should not be checked. Any value means `true`.

The configuration can also be provided by environment variables.

### Scaleway

This returns the first private IP address of all servers for the given
`region` with the given `tag_name`.

```shell-session
$ consul agent -retry-join "provider=scaleway organization=my-org tag_name=consul-server token=... region=..."
```

```json
{
  "retry_join": [
    "provider=scaleway organization=my-org tag_name=consul-server token=... region=..."
  ]
}
```

- `provider` (required) - the name of the provider ("scaleway" in this case).
- `region` (required) - the name of the region.
- `tag_name` (required) - the name of the tag to auto-join on.
- `organization` (required) - the organization access key to use for auth (equal to access key).
- `token` (required) - the token to use for auth.

### TencentCloud

This returns the first IP address of all servers for the given `region` with the given `tag_key` and `tag_value`.

```shell-session
$ consul agent -retry-join "provider=tencentcloud region=... tag_key=consul tag_value=... access_key_id=... access_key_secret=..."
```

```json
{
  "retry_join": [
    "provider=tencentcloud region=... tag_key=consul tag_value=... access_key_id=... access_key_secret=..."
  ]
}
```

- `provider` (required) - the name of the provider ("tencentcloud" in this case).
- `region` (required) - The TencentCloud region.
- `tag_key` (required) - The tag key to auto-join on.
- `tag_value` (required) - The tag value to auto-join on.
- `address_type` (optional) - "private_v4" or "public_v4", default is "private_v4".
- `access_key_id` (required) - The secret id of TencentCloud.
- `access_key_secret` (required) - The secret key of TencentCloud.

This required permission to 'cvm:DescribeInstances'.
It is recommended you make a dedicated key used to auto-join the Consul datacenter.

### Joyent Triton

This returns the first PrimaryIP addresses for all servers with the given `tag_key` and `tag_value`.

```shell-session
$ consul agent -retry-join "provider=triton account=testaccount url=https://us-sw-1.api.joyentcloud.com key_id=... tag_key=consul-role tag_value=server"
```

```json
{
  "retry_join": [
    "provider=triton account=testaccount url=https://us-sw-1.api.joyentcloud.com key_id=... tag_key=consul-role tag_value=server"
  ]
}
```

- `provider` (required) - the name of the provider ("triton" in this case).
- `account` (required) - the name of the account.
- `url` (required) - the URL of the Triton api endpoint to use.
- `key_id` (required) - the key id to use.
- `tag_key` (optional) - the instance tag key to use.
- `tag_value` (optional) - the tag value to use.

### vSphere

This returns the first private IP address of all servers for the given region with the given `tag_name` and `category_name`.

```shell-session
$ consul agent -retry-join "provider=vsphere category_name=consul-role tag_name=consul-server host=... user=... password=... insecure_ssl=[true|false]"
```

```json
{
  "retry-join": [
    "provider=vsphere category_name=consul-role tag_name=consul-server host=... user=... password=... insecure_ssl=[true|false]"
  ]
}
```

- `provider` (required) - the name of the provider ("vsphere" is the provider here)
- `tag_name` (required) - The name of the tag to look up.
- `category_name` (required) - The category of the tag to look up.
- `host` (required) - The host of the vSphere server to connect to.
- `user` (required) - The username to connect as.
- `password` (required) - The password of the user to connect to vSphere as.
- `insecure_ssl` (optional) - Whether or not to skip SSL certificate validation.
- `timeout` (optional) - Discovery context timeout (default: 10m)

### Packet

This returns the first private IP address (or the IP address of `address type`) of all servers with the given `project` and `auth_token`.

```shell-session
$ consul agent -retry-join "provider=packet auth_token=token project=uuid url=... address_type=..."
```

```json
{
  "retry-join": [
    "provider=packet auth_token=token project=uuid url=... address_type=..."
  ]
}
```

- `provider` (required) - the name of the provider ("packet" is the provider here)
- `project` (required) - the UUID of packet project
- `auth_token` (required) - the authentication token for packet
- `url` (optional) - a REST URL for packet
- `address_type` (optional) - the type of address to check for in this provider ("private_v4", "public_v4" or "public_v6". Defaults to "private_v4")

### Linode

This returns the first private IP address of all servers for the given `region` with the given `tag_name`.

```shell-session
$ consul agent -retry-join "provider=linode region=us-east tag_name=consul-server"
```

```json
{
  "retry-join": ["provider=linode region=us-east tag_name=consul-server"]
}
```

- `provider` (required) is `linode`
- `api_token` (required) - The Linode API token to use
- `region` (optional) - The Linode region to filter on
- `tag_name` (optional) - The tag name to filter on
- `address_type` (optional) - the type of address to check for in this provider ("private_v4", "public_v4" or "public_v6". Defaults to "private_v4")

Variables can also be provided by environment variables:

- `LINODE_TOKEN` for `api_token`

### Kubernetes (k8s)

The Kubernetes provider finds the IP addresses of pods with the matching
[label or field selector](https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/).
This is useful for non-Kubernetes agents that are joining a server cluster
running within Kubernetes.

The pod IP is used by default, which requires that the agent connecting can
network to the pod IP. The `host_network` boolean can be set to true to use
the host IP instead, but this requires the agent ports (Gossip, RPC, etc.)
to be exported to the host as well.

By default, no port is specified. This causes Consul to use the default
gossip port (default behavior with all join requests). The pod may specify
the `consul.hashicorp.com/auto-join-port` annotation to set the port. The value
may be an integer or a named port.

```shell-session
$ consul agent -retry-join "provider=k8s label_selector=\"app=consul,component=server\""
```

```json
{
  "retry-join": ["provider=k8s label_selector=..."]
}
```

- `provider` (required) - the name of the provider ("k8s" is the provider here)
- `kubeconfig` (optional) - path to the kubeconfig file. If this isn't
  set, then in-cluster auth will be attempted. If that fails, the default
  kubeconfig paths are tried (`$HOME/.kube/config`).
- `namespace` (optional) - the namespace to search for pods. If this isn't
  set, it defaults to all namespaces.
- `label_selector` (optional) - the label selector for matching pods.
- `field_selector` (optional) - the field selector for matching pods.

The Kubernetes token used by the provider needs to have permissions to list pods
in the desired namespace.