Skip to main content

​
Managing Instances

The Instances dashboard gives you a real-time overview of all your GPU instances and tools to manage them.

​
Dashboard View

The instances list shows the following for each instance:
ColumnDescription
NameInstance name (auto-generated or custom).
StatusCurrent lifecycle state (creating, deployed, stopped, terminated, etc.).
GPUGPU type and count (e.g., 1x RTX 4090).
NetworkUpload and download speed.
Hourly RateCost per hour for this instance.
CreatedDate and time the instance was deployed.

​
Instance Detail Page

Click on any instance to view its full details:
  • Specs — GPU model, GPU count, CPU cores, RAM, and disk size.
  • Network — Public IP address, SSH port, and bandwidth.
  • Region — The region where the instance is running.
  • SSH Details — Connection command ready to copy.
  • Mounted Storage — Any attached storage volumes.
  • Cost — Hourly rate and total spend for this instance.

​
Monitoring

After connecting via SSH, use these commands to monitor your instance:

​
GPU Usage

# One-time GPU status
nvidia-smi

# Continuous monitoring (updates every 1 second)
watch -n 1 nvidia-smi

​
CPU and Memory

# Interactive process viewer
htop

​
Disk Usage

# Check disk space
df -h

# Check directory sizes
du -sh /root/*

​
Terminating an Instance

Termination is permanent. All local disk data is deleted and cannot be recovered. Transfer important files or use a storage volume before terminating.

​
Steps to Terminate

  1. Go to Dashboard > Instances.
  2. Click on the instance you want to terminate.
  3. Click the Terminate button.
  4. Confirm the termination in the dialog.
Billing stops once the instance reaches the terminated state.

​
Best Practices

Use Storage Volumes

Attach persistent storage to keep datasets and model weights safe across instance lifecycles.

Enable Auto-Recharge

Prevent unexpected instance termination by enabling auto-recharge in your Billing settings.

Monitor Costs

Check the Billing page regularly to track your spend and set up usage alerts.

Save Work Before Terminating

Always download important files or push to a remote repository before terminating an instance.