Instance Downtime
- Confirmation Criteria: The instance status is “Shutdown - Unexpected Downtime.”
Instance Downtime Compensation
Instance Downtime Compensation
Calculation: Compensation duration is calculated from the last startup to the downtime moment, capped at 24 hours.Compensation Amount:
- For Pay-as-you-go instances:
- Compensation Voucher Amount = Duration (hours) * Instance Unit Price (USD/hour)
- For Subscription instances:
- Unit Price = Subscription Amount / Subscription Duration
- Compensation Voucher Amount = Duration (hours) * Unit Price
Network Failure
- Confirmation Criteria: Network anomalies in the instance (not caused by local user-to-instance connection issues, such as using a VPN).
Network Failure Compensation
Network Failure Compensation
Calculation: Total duration of the anomaly (in minutes) * 1.5, rounded up to the nearest hour.Compensation Amount:
- For Pay-as-you-go instances:
- Compensation Voucher Amount = Duration (hours) * Instance Unit Price (USD/hour)
- For Subscription instances:
- Unit Price = Subscription Amount / Subscription Duration
- Compensation Voucher Amount = Duration (hours) * Unit Price
Local Disk Failure
- The system disk and data disk (/root/gpuhub-tmp) provided by GPUhub are local disks on the host. While local disks offer better performance, they lack redundant backups and may experience disk failures.
- Yes. We strongly recommend that users regularly back up important data to prevent potential losses, again.
Other Scenarios
The following scenarios are not eligible for compensation:- The instance status is normal (“Running”), but the program exits prematurely due to reasons such as not running in the background.
- GPU memory not being correctly released due to driver issues, which can be resolved without restarting the instance.
- System dependency errors caused by user actions, resulting in inaccessible services like JupyterLab. These can be resolved by resetting the system.
- Other similar scenarios.