Multi-Machine Sessions¶
Sessions with multiple machines, created by Pool, Virtual Lab, and Cluster formations.
Overview¶
Multi-machine sessions are created by formation types that provision more than one instance. Each machine within the session can be managed independently (start, stop, connect, snapshot).
Session Structures¶
Machine Pool Sessions¶
Created by Machine Pool formations.
All machines are identical (same image, same instance type) and run independently. Each can be connected to, stopped, or snapshotted individually. Ideal for workshops where each participant gets their own machine.
Virtual Lab Sessions¶
Created by Virtual Lab formations.
The proxy machine is a RosettaHub container that provides browser-based remote desktop access. The target machine is your actual workstation running your chosen cloud image. You connect through the proxy -- no VPN, SSH client, or RDP client needed.
Big Data Cluster Sessions¶
Created by Big Data Cluster formations.
Session
├── Proxy Machine (browser gateway)
├── Master Node (cluster controller)
├── Worker Node 1
├── Worker Node 2
└── ...
The proxy provides browser access to the cluster. The master node runs the resource manager (YARN, etc.). Worker nodes execute data processing tasks. Worker count is configurable and can use Spot instances for cost savings.
HPC Cluster Sessions¶
Created by HPC Cluster formations.
Session
├── Head Node (job scheduler)
├── Compute Node 1
├── Compute Node 2
├── ...
└── Shared Storage
The head node runs the job scheduler (Slurm, etc.). Compute nodes execute submitted jobs. All nodes share a common file system for data access.
Managing Multi-Machine Sessions¶
Each machine within a multi-machine session supports the same lifecycle actions as single-machine sessions (start, stop, hibernate, reboot, delete). You can manage machines individually or use batch actions:
| Action | Description |
|---|---|
| Stop All | Stop all machines in the session |
| Terminate All | Delete all machines in the session |
| Start (individual) | Resume a specific stopped machine |
| Connect (individual) | Connect to a specific machine |
Spot Instances in Multi-Machine Sessions¶
Worker and compute nodes in cluster and pool sessions can use Spot/Preemptible instances for 60-90% cost savings. If a Spot instance is reclaimed:
- Hibernation preserves machine state where supported
- Other machines in the session continue running
- The session remains active with reduced capacity
Related Topics¶
- Sessions Overview -- All session types, states, and best practices
- Machine Pool Formations -- Create pool sessions
- Virtual Lab Formations -- Create virtual lab sessions
- Big Data Clusters -- Create cluster sessions
- HPC Clusters -- Create HPC sessions