Known Issues

Current limitations and known issues in Kaisar AI Ops.

[!NOTE] This page is regularly updated. Last updated: 2024-01-15

Platform Issues

Cluster and Jobs Sections Not Accessible

Status: Under Development Severity: Low Description: The "Cluster" and "Jobs" menu items are visible in the sidebar but not currently clickable or functional. Workaround: Use the Experiments feature for job submission. ETA: Q2 2024

Large Dataset Upload Timeout

Status: Known Issue Severity: Medium Description: Uploading datasets larger than 10GB may timeout in the browser. Workaround: Use the CLI or API for large uploads, or split datasets into smaller chunks. Fix: Planned for next release

Dashboard Refresh Delay

Status: Known Issue Severity: Low Description: Dashboard metrics may take up to 60 seconds to refresh. Workaround: Manually refresh the page for latest data. Fix: Under investigation

Experiments

GPU Memory Not Released

Status: Known Issue Severity: Medium Description: In some cases, GPU memory is not fully released after experiment completion. Workaround: Restart the experiment or wait for automatic cleanup (5 minutes). Fix: Planned for v1.5

Experiment Logs Truncated

Status: Known Issue Severity: Low Description: Experiment logs longer than 10MB may be truncated in the UI. Workaround: Download full logs via API or CLI. Fix: Improved log viewer in development

Concurrent Experiment Limit

Status: By Design Severity: Low Description: Maximum 50 concurrent experiments per organization. Workaround: Queue additional experiments or request quota increase. Note: This is a configurable limit

Models

ONNX Model Preview Not Supported

Status: Feature Request Severity: Low Description: ONNX models cannot be previewed in the UI. Workaround: Download and inspect locally. ETA: Q3 2024

Model Download Speed

Status: Known Issue Severity: Low Description: Large model downloads (>5GB) may be slow. Workaround: Use direct S3/GCS access if configured. Fix: CDN integration planned

Deployments

Cold Start Latency

Status: Known Issue Severity: Medium Description: First request to a deployment may take 30-60 seconds (cold start). Workaround: Configure minimum instances > 0 to keep instances warm. Fix: Improved cold start optimization in progress

Auto-scaling Delay

Status: By Design Severity: Low Description: Auto-scaling takes 2-3 minutes to provision new instances. Workaround: Set appropriate min/max instances based on expected traffic. Note: This is normal cloud provider behavior

API

Rate Limit Headers Missing

Status: Known Issue Severity: Low Description: Some API endpoints don't return rate limit headers. Workaround: Monitor 429 responses. Fix: Planned for v1.4

Webhook Retry Logic

Status: Known Issue Severity: Medium Description: Failed webhooks are retried only 3 times. Workaround: Implement your own retry logic or use a message queue. Fix: Configurable retry policy in development

Browser Compatibility

Safari Private Mode Issues

Status: Known Issue Severity: Low Description: Some features may not work in Safari Private Mode due to cookie restrictions. Workaround: Use normal browsing mode or a different browser. Fix: Under investigation

Internet Explorer Not Supported

Status: By Design Severity: N/A Description: Internet Explorer is not supported. Workaround: Use Chrome, Firefox, Safari, or Edge. Note: No plans to support IE

Integrations

GitHub Enterprise Server

Status: Limited Support Severity: Medium Description: GitHub Enterprise Server integration has limited functionality. Workaround: Use GitHub.com or manual Git integration. Fix: Full support planned for Q2 2024

Slack Notification Delays

Status: Known Issue Severity: Low Description: Slack notifications may be delayed by 1-2 minutes. Workaround: Use email for time-critical notifications. Fix: Under investigation

Performance

Large Organization Slowdown

Status: Known Issue Severity: Medium Description: Organizations with >10,000 experiments may experience slow dashboard loading. Workaround: Use filters and pagination, or archive old experiments. Fix: Database optimization in progress

Search Performance

Status: Known Issue Severity: Low Description: Search may be slow for organizations with many resources. Workaround: Use specific filters to narrow results. Fix: Search index optimization planned

Security

MFA Recovery Codes

Status: Known Issue Severity: Medium Description: Recovery codes are shown only once and cannot be regenerated. Workaround: Save recovery codes securely when first generated. Fix: Recovery code regeneration feature planned

Reporting Issues

If you encounter an issue not listed here:

1. Check the FAQ

2. Review Troubleshooting

3. Search existing issues

4. Submit a new issue via Support

When reporting, include:

• Steps to reproduce

• Expected vs. actual behavior

• Error messages and logs

• Browser/OS information

• Screenshots if applicable

Workaround Index

Quick reference for common workarounds:

Issue	Workaround
Large dataset upload	Use CLI/API
GPU memory not released	Wait 5 minutes or restart
Slow dashboard	Use filters and pagination
Cold start latency	Set min instances > 0
Webhook failures	Implement retry logic
Safari private mode	Use normal mode

Next Steps

• Review FAQ for common questions

• Check Troubleshooting for solutions

• Contact Support for assistance