Troubleshooting & FAQ
Troubleshooting
Common Issues
| Issue | Solution |
|---|---|
| "Access Denied" when using CLI tools | Go to your dataset's Setup tab to view your access keys. Copy the Access Key and Secret Key, then reconfigure your CLI tool (s3cmd, mc, etc.). |
| Dataset not visible in notebook | The dataset wasn't attached after launch. Go to your running instance and attach the dataset from the storage options. |
| Cannot find uploaded files | Files may take a few seconds to appear. Refresh the Data Objects tab or list the bucket contents to verify the upload completed successfully. |
| Slow upload speeds | For large files, use the MinIO CLI (mc cp) or s3cmd instead of the web UI. Uploading compressed archives is faster than uploading many small files individually. |
Security Overview
Authentication
- Access Keys: Standard S3-style Access Key and Secret Key pairs (EOS Only).
- RBAC: Project-level access controls determine who can view or delete datasets.
Data Security
- Encryption at Rest: All data in EOS is encrypted.
- Server-Side Encryption (SSE-S3): Managed by TIR.
- Server-Side Encryption with Customer Keys (SSE-C): Managed by you.
- Encryption in Transit: All traffic is secured via TLS/SSL (HTTPS).
Network Isolation
- Datasets are logically isolated within your project.
- Public access is blocked by default; access is only possible via authorized credentials or valid internal mounts.
Frequently Asked Questions (FAQ)
General Questions
Q: What's the difference between EOS Bucket and Disk storage?
A: EOS Bucket is S3-compatible object storage with virtually unlimited scalability, ideal for storing large datasets, backups, and model artifacts. Disk is PVC-backed block storage with fixed size, offering low latency and high IOPS, best suited for temporary scratch space and high-speed caching.
Q: Can I change the storage type after creating a dataset?
A: No, you cannot change the storage type (EOS Bucket or Disk) after a dataset is created. You'll need to create a new dataset with the desired storage type and migrate your data.
Q: How do I access my dataset from a notebook or training job?
A: When launching a notebook or training job, simply select the datasets you want to mount. They will automatically appear under the /datasets/<dataset-name> directory.
Storage & Capacity
Q: What's the maximum size for a dataset?
A: EOS Bucket storage is virtually unlimited. Disk storage has a fixed size that you specify during creation, but it can be increased later if needed.
Q: How do I increase the size of my Disk dataset?
A: You can increase Disk storage size through the resize disk option in the dataset home page. Note that you cannot decrease the size once it's been increased.
Encryption & Security
Q: Should I use E2E Managed or User-Managed encryption?
A: Use E2E Managed encryption for ease of use and hassle-free key management. Choose User-Managed encryption only if you have strict compliance requirements for key ownership. Remember with User-Managed encryption, key loss means permanent data loss.
Q: Where can I find my dataset access keys?
A: Access keys are provided in the Setup section of your dataset when you create it. You can view them anytime by navigating to your dataset and clicking on the Setup tab.
Q: Is encryption free of cost?
A: Yes, both E2E Managed and User-Managed encryption are provided free of cost. There are no additional charges for encrypting your data at rest.
Data Management
Q: How do I delete data from my dataset?
A: For EOS Buckets, use the web UI's Data Objects tab or the MinIO CLI (mc rm command). For Disk storage, access the mounted directory and delete files normally.
Q: What happens to my data if I delete a dataset?
A: All data in the dataset is permanently deleted and cannot be recovered. Make sure to backup any important data before deleting a dataset.
Q: Can I share a dataset with other team members?
A: Yes, datasets are accessible to all members of your TIR project. You can share access keys with team members, but follow the principle of least privilege for security.
Lifecycle Rules
Q: What are lifecycle rules and when should I use them?
A: Lifecycle rules automatically delete objects after a specified number of days. Use them to manage storage costs by automatically cleaning up temporary data, old logs, or outdated experiment results.
Q: Can I recover data deleted by lifecycle rules?
A: No, lifecycle deletion is irreversible. Objects deleted by lifecycle rules cannot be recovered.
Q: Can I have multiple lifecycle rules on the same dataset?
A: Yes, you can create multiple lifecycle rules with different prefixes to manage different types of data with different retention periods.