---
title: Clone Instance
---

import { CloneInstanceNav } from './CloneInstanceCards'

# Clone Instance

<CloneInstanceNav />

The **Clone Instance** action lets you spin up a new instance using the same image, version, and compute configuration as an existing instance — without manually re-selecting each option in the create flow.

---

## 1. Overview

Clicking **Clone Instance** takes you to a pre-filled **Configure** page for the same image as the source instance, with the compute resource type and configuration already selected to match. You can review, adjust, and launch it like a normal instance creation.

| Detail | Description |
|--------|-------------|
| **Where to find it** | **Actions** menu (⋮) on the instance row in **Manage Instances**, or on the instance details page. |
| **Result** | A new, independent instance — it does not modify or affect the source instance. |

---

## 2. When to Use

| Scenario | Description |
|----------|-------------|
| **Scale out identical workers** | Launch additional instances with the same framework and hardware plan for parallel jobs, without repeating the create-instance steps. |
| **Recreate a working setup** | Quickly reproduce an instance's image, version, and compute configuration for a new task or teammate. |
| **Test before scaling** | Spin up a duplicate configuration to test changes without touching the original instance. |

---

## How Clone Instance Works

Clone Instance reuses the standard create-instance flow, pre-filling it with the source instance's:

- Image, version, and app
- Compute resource type (GPU, CPU, Private Cluster, or Spot Instances) and plan / SKU
- Workspace (disk) size and storage encryption setting
- SSH access toggle and attached SSH key(s)
- Start scripts and add-ons
- Attached Datasets (EOS-backed), Shared File System (SFS), and Parallel File System (PFS) mounts, including their mount paths
- Placement / affinity group, for Private Cluster instances

The instance name is **not** reused — a new, randomly generated name (for example, `clone-instance-<id>`) is pre-filled, which you can edit before launching.

:::info
Clone Instance only copies **configuration**, not the running environment or workspace data. See [What Carries Over](#what-carries-over) below.
:::

---

## Step-by-Step Process

1. Navigate to **Instances**, open the instance you want to clone, and go to its **Instance Details** page.
2. Click the **⋮ (Actions)** menu and select **Clone Instance**.
3. On the **Configure** page that opens, review the pre-filled **Instance Name**, image version, computing resource type, and configuration.
4. Adjust any field as needed — instance name, compute plan, storage, network and security, or advanced settings — the same way you would when [creating a new instance](/docs/tir/Nodes/create-instance).
5. Click **Launch** (or **Create**) to provision the new instance.

---

## What Carries Over

| Carried Over | Not Carried Over |
|--------------|-------------------|
| Image, version, and app | Instance name (a new name is auto-generated) |
| Compute resource type and plan / SKU (GPU / CPU / Private Cluster / Spot) | Security Group (left blank — not even the usual default SSH security group is auto-applied) |
| Workspace size and storage encryption setting | Reserved IP (always off on the clone, regardless of the source) |
| SSH access toggle and attached SSH key(s) | Installed packages and files created during the session |
| Start scripts and add-ons | Actual workspace file contents (`/home/jovyan`) |
| Attached Datasets (EOS), SFS, and PFS mounts | Dataset along with its Disk (PVC) |

:::warning
The cloned instance starts with a fresh workspace disk of the same size — it does not copy the actual files inside `/home/jovyan` or any packages installed after the image was built. If you need that data available, save an image of the source instance first and select it while configuring the clone.
:::

---