Launch a Machine from an Image¶
Overview¶
In this tutorial, you'll learn how to launch a cloud machine instance directly from an image registered in the RosettaHub Supercloud platform. When you launch an image, RosettaHub automatically creates a formation and session, provisions the cloud instance, and provides you with connection details -- all within your organization's budget controls and policy enforcement.
You can launch instances as spot machines (discounted instances that may be interrupted by the cloud provider) or on-demand machines (guaranteed instances at standard rates). Spot machines are ideal for development, testing, and fault-tolerant workloads.
Prerequisites¶
- [ ] RosettaHub account with active subscription
- [ ] At least one image registered in your Images panel (see Register a Machine Image)
- [ ] A Cloud Key associated with the image (see Working with Keys)
Steps¶
Step 1: Select an Image to Launch¶
From the dashboard, scroll to the Images panel. Browse or search for the image you want to launch, and click on it to select it.
Tip
If you cannot find the image you need, you can register a new one by following the Register a Machine Image tutorial.
Step 2: Choose Spot or On-Demand¶
The launch dialog opens with Spot Machine selected by default. Choose the instance pricing model:
- Spot Machine - Launch a spot/preemptible instance at a discounted rate. The cloud provider may reclaim this instance if demand increases.
- Machine - Launch an on-demand instance with guaranteed availability at standard rates.
Tip
Spot instances offer significant cost savings for development, experimentation, and batch workloads. Use on-demand for production or long-running critical tasks.
Step 3: Enter a Label¶
Input a descriptive label for your machine. This name will appear in the Machines panel and help you identify the instance.
Step 4: Verify the Keys¶
The Keys field is pre-populated with the Cloud Key associated with the image. Leave this unchanged unless you need to deploy to a different region or cloud account.
Step 5: Configure Root Volume Size¶
The root volume size defaults to the minimum required for the image. Increase this if your workload needs additional disk space.
Note
The root volume stores the operating system and any data written to the instance's local disk. For large datasets, consider attaching separate storage volumes via Storages.
Step 6: Select the Instance Type¶
Choose an instance type that matches your compute requirements (e.g., p2.xlarge for GPU workloads, m5.large for general purpose).
For spot machines, the current market price is displayed next to each instance type to help you estimate costs.
Step 7: Review Bid Price (Spot Only)¶
If launching a spot machine, the Maximum bid price is set to the current market price by default. Leave this unchanged unless you want to cap your maximum spend.
Step 8: Confirm Shutdown Mode¶
The default shutdown mode is manual shutdown, meaning you are responsible for stopping or deleting the machine when you are done. Leave this selected unless your organization has configured automated shutdown policies.
Warning
With manual shutdown selected, the machine will continue running (and incurring costs) until you explicitly stop or delete it. Always shut down machines when you are finished working.
Step 9: Launch the Machine¶
Click Launch to provision the instance.
RosettaHub automatically creates a formation and a session for this launch. The session appears in your dashboard.
Step 10: Monitor the Session¶
Watch the session status in the dashboard:
- Pending - The spot or on-demand request has been submitted
- Creating - The cloud provider is provisioning the instance
- Running - The machine is ready to use
For spot machines, a green tick appears on the session once the spot request is fulfilled.
Step 11: Connect to Your Machine¶
Once the machine is running, right-click the session and select Get Connectivity Info. A window opens with connection instructions.
Connection methods depend on the operating system:
| OS | Connection Methods |
|---|---|
| Linux | SSH, Web Terminal, VNC (if configured) |
| Windows | RDP, Web RDP |
You can also click the machine in the Machines panel to access web-based interfaces (Jupyter, RStudio, etc.) if the image includes them.
Key Concepts¶
| Term | Description |
|---|---|
| Spot Machine | Discounted instance that may be interrupted by the cloud provider |
| On-Demand Machine | Guaranteed instance at standard rates |
| Formation | Automatically created recipe/template for the launch |
| Session | Active deployment containing the running machine |
Next Steps¶
- Launching Your First Formation - Launch multi-machine deployments
- Sessions Guide - Managing running sessions with real-time cost tracking
- Share a Machine Image - Let others launch from your images
- Images User Guide - Complete images documentation
- Cloud Operations - Governance, budgets, and policy enforcement
Troubleshooting¶
Spot request not fulfilled -- no green tick appears
Spot capacity is subject to cloud provider availability. This may indicate:
- The requested instance type has no spot capacity in your region
- Your bid price is below the current market price
- The cloud provider has limited spot availability
Try selecting a different instance type or region, or switch to an on-demand machine.
Machine stuck in 'Creating' status
This may indicate:
- Cloud provider quota limits for your account
- Region availability issues
- Budget limits exceeded in your organization
Check your cloud account status in Cloud Operations or contact your administrator.
Cannot connect to the machine after launch
Ensure that:
- The machine status shows Running (not still creating or pending)
- The image has the correct username configured (see Register a Machine Image)
- The required ports are open (SSH port 22 for Linux, RDP port 3389 for Windows)
- Your network allows outbound connections to the cloud provider
Launch fails with a permission or budget error
Your account may lack:
- Required instance type permissions
- Access to the selected region
- Sufficient budget allocation
Contact your organization administrator or check your Cloud Accounts settings.