---
title: "ezgh compute autoscaling-groups create"
description: "Create an autoscaling group."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.ezghcloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# ezgh compute autoscaling-groups create

{/* Generated by ezgh-cli's internal/tools/clidocs from the ezgh command tree. Don't edit: change the command's help in ezgh-cli and regenerate. */}

```sh
ezgh compute autoscaling-groups create <name> --min n --max n --instance-type t --image <alias|img_id> [--desired n] [launch flags] [policy flags] [flags]
```

Create an autoscaling group: `--min` and `--max` set its size (at most 100 instances), `--desired` where it starts (default `--min`), and the launch flags what each member launches, as `ezgh compute instances launch` takes them (a group's members can't use existing volumes or interfaces). Members are named &lt;group&gt;-&lt;random&gt;. The group scales to hold its policies' targets: `--cpu-target` and `--memory-target` are percentages of the members' average CPU and memory utilization; `--promql-policy` name=&lt;name&gt;,target=&lt;target&gt;,query=&lt;query&gt; (repeat) holds a PromQL query that gives one series at its target, with \_\_asg_instances\_\_ standing for the group's warm members' instance IDs. Without a policy the group stays at `--desired`. `--target-group` (repeat, at most 5) registers each member in a load balancer target group (tgp\_…) once it's running, and drains it from there before it goes; each one needs lb.targetGroups.registerTargets and lb.targetGroups.deregisterTargets.

It prints the CreateAutoscalingGroup operation, or with `--wait` waits until the group is active and prints it. Needs compute.autoscalingGroups.create.

Part of [`ezgh compute autoscaling-groups`](/cli/reference/compute-autoscaling-groups).

## Examples

```sh
ezgh compute autoscaling-groups create web --min 2 --max 10 --instance-type gp-1.large --image linux/ubuntu:26.04 --ssh-key laptop --cpu-target 60 --wait
```

## Flags

| Flag | Type | Default | Description |
| --- | --- | --- | --- |
| `--batch-percent` | int |  | the share of the group a release replaces at a time, 1–100 (default 25) |
| `--bot` | string |  | attach a Bot (bot\_…): code on the instance gets short-lived credentials for it |
| `--cpu-target` | string |  | keep the members' average CPU utilization at this percentage (policy cpu) |
| `--desired` | int |  | how many instances the group aims for, between `--min` and `--max` (default on create: `--min`) |
| `--firewall` | string, repeatable |  | a firewall for the new interface (ID, UUID or name), in place of the project's default (repeat) |
| `--health-check-grace` | int |  | seconds before a new member's health is judged (default 300) |
| `--image` | string |  | the image alias, such as linux/ubuntu:26.04, or an exact image ID (img\_…) |
| `--instance-type` | string |  | the instance type, such as gp-1.large (required); ezgh compute instance-types list shows the organization's |
| `--instance-warmup` | int |  | seconds before a new member's metrics count (default 300) |
| `--ipv4` | boolean |  | give the new interface a public IPv4 address |
| `--ipv6` | boolean |  | with `--primary-subnet`: give it a public IPv6 /64 |
| `--max` | int |  | the most instances the group runs (at most 100) |
| `--memory-target` | string |  | keep the members' average memory utilization at this percentage (policy memory) |
| `--min` | int |  | the fewest instances the group keeps (0–100) |
| `--new-volume` | string, repeatable |  | create and attach a volume: &lt;GiB&gt;\[:&lt;volume type&gt;\] (repeat) |
| `--no-firewalls` | boolean |  | no firewall on the new interface |
| `--no-ipv4` | boolean |  | with `--primary-subnet`: no public IPv4 address, whatever the subnet's auto-assign |
| `--no-ipv6` | boolean |  | with `--primary-subnet`: no public IPv6 /64, whatever the subnet's auto-assign |
| `--primary-subnet` | string |  | make the new interface private, in this VPC subnet (ID, UUID or name) |
| `--promql-policy` | string, repeatable |  | a policy on a PromQL query: name=&lt;name&gt;,target=&lt;target&gt;,query=&lt;query&gt;; the query comes last and takes the rest of the value (repeat) |
| `--root-iops` | int |  | the root volume's provisioned IOPS, for types that take it |
| `--root-size` | int |  | the root volume's size in GiB (default: the image's minimum) |
| `--root-throughput` | int |  | the root volume's provisioned throughput in MB/s, for types that take it |
| `--root-volume-type` | string |  | the root volume's type (default: the block store's default) |
| `--scale-in-cooldown` | int |  | seconds a smaller size must hold before the group scales in (default 300) |
| `--ssh-key` | string, repeatable |  | an SSH key of the project to install, by ID, UUID or name (repeat, at most 10) |
| `--subnet` | string, repeatable |  | add a new private interface in this VPC subnet (ID, UUID or name), after the primary (repeat) |
| `--tag` | string, repeatable |  | a tag, key=value (repeat) |
| `--target-group` | string, repeatable |  | a target group (tgp\_…) the members are registered in, at most 5; repeat for more |
| `--user-data` | string |  | a file of user-data (cloud-init), - for stdin, at most 16 KiB |
| `--wait` | boolean |  | wait until the group is active, and print it then |
| `--wait-timeout` | duration | `10m0s` | how long `--wait` waits |

This command also takes the [global flags](/cli/reference#global-flags).

Source: https://docs.ezghcloud.com/cli/reference/compute-autoscaling-groups-create/index.mdx
