Optional Platform Toolbox¶
Overview¶
Lists common additions for a small k3s GitOps platform. Install only when a concrete requirement exists; every controller consumes memory and expands the upgrade and security surface.
Recommended Tools¶
| Need | Tool | Guidance |
|---|---|---|
| GitOps visualization | Argo CD UI | Primary view of desired versus live state |
| Cluster terminal UI | k9s | Run on the operator machine; nothing to install in cluster |
| Database UI | CloudBeaver or pgAdmin | Optional; port-forward and disable when unused |
| Kafka UI | Redpanda Console or AKHQ | Optional; internal-only access |
| Redis UI | RedisInsight | Prefer a local client through port-forward |
| Backups | Velero plus database-native backup | Store backups outside the VPS |
| S3-compatible storage | External provider or MinIO | External storage is safer; single-node MinIO shares the VPS failure domain |
| DNS automation | external-dns | Use only with a supported provider and tightly scoped API token |
| Image automation | Renovate | Open pull requests for pinned chart and image updates |
| Policy | Kyverno | Add after baseline GitOps is stable |
| Runtime troubleshooting | kubectl debug and ephemeral containers |
Avoid debug tools inside production images |
| External uptime checks | Uptime Kuma or a hosted monitor | A hosted monitor can detect total VPS failure |
| Progressive delivery | Argo Rollouts | Add canary or blue-green deployment after basic delivery is stable |
| Container registry | GHCR, Harbor | Prefer a hosted registry; Harbor is heavy for a small VPS |
Baseline Selection¶
For an 8 GB VPS, begin with Argo CD, cert-manager, PostgreSQL, Redis, and metrics/Grafana. Add Loki and a broker only after checking sustained memory and disk headroom. Keep database and Kafka UIs off by default.
Operational Rules¶
- Pin versions; let Renovate propose upgrades.
- Define requests, limits, PVC sizes, and retention for every stateful component.
- Keep admin UIs private and credentials in the chosen secret manager.
- Send backups to a different provider and failure domain; test restores.
- Maintain a resource budget so optional tools cannot evict core workloads.