Frequently Asked Questions

Use Troubleshooting for a failing deployment and lablink doctor to check CLI prerequisites.

Deployment

Do I need AWS?

Only for provider: aws. With provider: manual, the allocator runs in Docker on a machine you control and you register client boxes yourself. See Bring-Your-Own Clients.

Can I deploy without GitHub Actions?

Yes. The recommended AWS path is lablink configure followed by lablink deploy. If you own a template repository, you can also run its OpenTofu files locally with the current deployment_name and environment inputs; see Manual OpenTofu Deployment.

What triggers a template-repository deployment?

Deploy LabLink Infrastructure runs for a push to test, manual dispatch with deployment_name and environment, or its configured repository dispatch. A push to main alone does not deploy. The workflow reads allocator.image_tag from config.yaml; it has no image-tag input. See Template Repository Deployment.

Can I use another AWS region?

Yes. The allocator AMI resolves per region. Set app.region and either leave machine.ami_id empty to resolve the default Deep Learning Base client AMI in that region, or provide a compatible AMI ID. lablink doctor checks that the selected image exists there. See Configuration.

Can I use my own image or software?

Yes. machine.image selects the client Docker image and machine.ami_id selects its AWS AMI on either AWS deployment path. The client AMI must have Docker, an NVIDIA driver, and the NVIDIA container runtime for GPU workloads. See Adapting LabLink.

Access and security

How do I set passwords?

The CLI prompts or uses saved values according to provider. A template repository keeps password placeholders in the committed config and replaces them from GitHub secrets during its workflow. See Change Default Passwords.

Why does the browser show 'Not Secure'?

ssl.provider: none serves HTTP. Choose an HTTPS mode in Configuration before public use. An HTTP viewer also loses H.264 streaming and falls back to JPEG/WebP.

Can I change SSL mode later?

Yes. Change ssl.provider and its required DNS or certificate settings, then redeploy through the same path. Use lablink deploy for the CLI or Deploy LabLink Infrastructure for a template repository. See DNS Configuration.

Are allocator and client SSH keys the same?

No. The deployment state holds the allocator key. The allocator's own OpenTofu workspace generates the client-VM keypair; its output is lablink_private_key_pem. See SSH Access.

Operations

How do I create client VMs?

Use Create New VM Instance in the admin dashboard or lablink client launch --num-vms N. The allocator's POST /api/launch takes num_vms, returns 202 with a job_id, and exposes progress at /api/operations. See Workshop Guide.

What happens when I destroy the allocator?

lablink destroy on AWS destroys the client fleet as part of the CLI workflow. The template's Destroy LabLink Infrastructure workflow attempts client-VM teardown from S3 state before removing the allocator. Use the same deployment name and environment used for apply. See Destroying a Deployment.

Do stopped instances or Elastic IPs cost money?

A stopped EC2 instance stops its compute usage charge, while attached EBS storage still accrues charges. AWS bills public IPv4 addresses whether associated or idle. See Cost Estimation.

Help

Where can I report a bug or ask a question?

Search GitHub Issues, then open a new issue with the commands you ran, the result, relevant logs, and your environment. The repository does not have Discussions enabled.