Maintenance and Troubleshooting

Since hardware failures and other issues are unavoidable, unforeseen circumstances may arise due to force majeure. All compensation will be issued to the account in the form of a voucher of equal value with no minimum spending requirement. The following handling and compensation measures apply depending on the specific situation:

Instance Outage

Eligibility Criteria: The instance status must be “已关机-意外宕机” and the instance must be unable to boot on its own. (Unexpected crashes caused by out-of-memory (OOM) errors in applications, rather than hardware issues, are not covered by compensation.)

Compensation Standards: The compensation duration is calculated from the time of the last boot-up to the time of the outage, with a maximum of 24 hours. The compensation voucher amount = duration (hours) × instance unit price (yuan/hour). For prepaid plans, the unit price is calculated by dividing the prepaid plan amount by the duration of the plan to determine the hourly rate.

Network Issues

Verification Criteria: Network issues with the instance (not caused by the connection from the user’s local network to the instance, such as a VPN connection causing network abnormalities)

Compensation Standard: Total duration of the incident (minutes) × 2, rounded up to the nearest hour. Compensation voucher amount = duration (hours) × instance unit price (yuan/hour); for subscription plans, the unit price is calculated as the subscription amount divided by the subscription duration.

Local Disk Failure

The system disk and data disk (/root/webcal-tmp) provided by WebCal for instances are local disks on the host. While local disks offer good performance, they lack redundant copies and are subject to the risk of disk failure(though the probability is low—enterprise-grade solid-state drives typically have a lifespan of over 10 years under normal read/write conditions). WebCal does not provide any reliability guarantees or services for this type of local storage. Therefore, please back up important data regularly during use (we recommend backing up to a local drive or WebCal file storage, etc.) to prevent loss in the event of an unexpected incident.

If all instance data is corrupted or lost due to host failure, hard drive failure, or other similar causes, the platform will reimburse all historical charges for that instance.

Other Scenarios

The following situations are not covered by the warranty:

1. The instance status is normal (running), but the program exited prematurely because it was not set to run in the background, etc.
2. Exceptions caused by user programs, such as failure to properly release GPU memory or RAM, that can be resolved without restarting the instance.
3. Exceptions caused by factors such as driver issues or user program memory overflows that can be resolved by restarting the instance; these situations fall within the normal stability range of the host, GPU, and operating system.
4. System dependency errors caused by user actions may ultimately result in JupyterLab and other services becoming inaccessible. In such cases, resetting the system can restore normal operation.
5. In the event of an unexpected shutdown caused by an out-of-memory (OOM) condition—which is not a hardware failure—users may restart the instance on their own.
6. If you find that your pay-as-you-go instance is in maintenance mode when you start it, simply wait until maintenance is complete to resume use. If your instance is about to be released, please contact customer service to request an extension.
7. Other exceptions caused by factors unrelated to hardware or the platform.