Deployments
Follow job logs, verify health, and recover when a deploy fails.
A deploy is a job that takes a service onto a server. The dashboard streams logs until the job succeeds or fails.
Watch a deploy
After you click Deploy:
- Open the deployment / job view for that service.
- Follow the log stream (image pull, container start, TLS, health).
- Wait until the job is marked complete.
If something fails, the log line that first errors is usually enough to fix DNS, disk, or a bad env var.
After success
- Confirm the health check is passing
- Open the public URL (custom domain) or the host and port documented for that stack
- For databases, connect from your app using the credentials Kubeara showed at deploy time — store them in your secret manager, not in chat
Updates and restarts
From the same service you can typically:
- Restart a container
- Update / redeploy when the template or image changes
- Inspect metrics (including GPU if present)
Routine restarts should not require SSH. Use SSH for host-level debugging (kernel, disk, NVIDIA driver), not for everyday deploys.
Common failures
| Failure | Likely cause |
|---|---|
| Image pull timeout | Outbound HTTPS from the server to the registry |
| TLS / certificate | DNS not pointing at the host; port 80/443 blocked |
| Health check loop | App not listening on the expected port; bad env |
| Disk full | Images and volumes filled the disk |
| GPU / VRAM | Model larger than free VRAM — check metrics first |
Automation
- MCP — ask Claude or Cursor to deploy or read logs
Status
Platform incidents: status.kubeara.dev.