Licensed to be used in conjunction with basebox, only.
// installation
Operations
Applies to
Product: Server · Audience: Platform Operator
basebox Server in production – whether you operate it yourself or basebox takes it on contractually. This section is the runbook: what to watch, where logs live, what gets backed up, how updates run, when to scale and how to diagnose problems.
Who operates
| Self-managed server | Operated by basebox | |
|---|---|---|
| These pages apply to | Your IT | basebox – and you as the client who knows what is happening |
| Access | Your team on the server | PAM/VPN, time-limited, approved by you |
| What stays with you | Everything | Application administration, data, approvals |
| Page | Customer-managed operations | Operation by basebox |
Buying the software does not include operations; the split is under Responsibilities.
The runbooks
| Topic | Question the page answers | Page |
|---|---|---|
| Monitoring | What do I need to watch, with what, and when to alert? | Monitoring |
| Logging | Where are the logs, what is in them (and what is not), how do I forward them? | Logging |
| Backup & restore | What do I back up, how often, and how do I restore? | Backup & restore |
| Updates | How do I update basebox safely – and how do I get back? | Updates |
| Scaling | When do I need more GPU, more nodes, more replicas? | Scaling |
| Troubleshooting | Symptom → cause → solution | Troubleshooting · FAQ |
Operating rhythm
Daily (automated, with alerts): pod state and restarts, health endpoints, GPU utilisation and memory, inference latency, free space on PVCs, backup success.
Weekly: review events and warnings, check database sizes and media volume, certificate expiry, read basebox release notes.
Monthly: rehearse a restore from a backup, operating system security updates, check capacity against growth, update the manifest (versions).
Per release: read the changelog, back up, render, upgrade in a maintenance window, acceptance check – see Updates.
Principles
- No silent fallback. If inference fails, basebox shows the error; monitoring the inference endpoint is therefore mandatory.
- Back up before every upgrade. Database migrations are forward-only.
- No automatic updates. Security and feature updates are announced by e-mail and release notes; installation is up to the operator.
- Logs without secrets.
infoin production; prompt content only attrace, credentials only atdebugin targeted diagnostic runs. - Maintain the manifest. Chart, app, driver, CUDA, Kubernetes, GPU Operator, image digests – the basis of every support case.
What does not belong in this section
Application administration (users, apps, connectors, policies) is under Administration. Security properties and compliance under Security & Compliance → Server. Support contact for operational questions: support@basebox.ai; security incidents: datenschutz@basebox.ai.
Next step: Customer-managed operations or Operation by basebox