sascha.dodenhoeft/expertfab-infra
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
dd714e14843250335d7e1893be415988b33fb714
expertfab-infra/docs/storage.md
Sascha Dodenhöft 2f8fb1349c Add storage flow & HA analysis diagram 
Graphviz DOT-based diagram showing the complete storage path:
App pods → PVCs (StorageClass/Retain-Policy) → Longhorn replicas →
K3s worker nodes → Proxmox hosts.

HA analysis annotated with color coding:
- Red: SPOF (control-plane on n01, share-manager on n01)
- Orange: Degraded on failure (CSI controllers on n02, RWX volumes)
- Green: HA covered (2 Longhorn replicas on different Proxmox hosts)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-03 14:00:45 +02:00

4.0 KiB
Raw Blame History

ExpertFab – Longhorn Storage Architektur

Speicherfluss & HA-Analyse: ../diagrams/storage_flow.png
(Quelle: ../diagrams/storage_flow.dot)

Übersicht (alt): ../diagrams/storage_architecture.png

StorageClasses

Name Reclaim Binding Verwendung
longhorn Delete Immediate RabbitMQ, Zitadel
longhorn-erpnext Retain Immediate ERPNext (Daten bleiben!)
longhorn-paperless Retain Immediate Paperless (Daten bleiben!)
longhorn-static Delete Immediate Manuell provisionierte Volumes
local-path Delete WaitForFirstConsumer Rancher local-path

Retain = PV bleibt erhalten wenn PVC gelöscht wird → Schutz vor Datenverlust

PVCs nach Namespace

namespace: erpnext

PVC Größe Mode StorageClass
data-erpnext-mariadb-sts-0 3 Gi RWO longhorn-erpnext
erpnext 3 Gi RWX longhorn-erpnext
erpnext-logs 1 Gi RWX longhorn-erpnext

namespace: paperless

PVC Größe Mode StorageClass
paperless-media 10 Gi RWO longhorn-paperless
paperless-consume 5 Gi RWO longhorn-paperless
paperless-data 5 Gi RWO longhorn-paperless
postgres-data 5 Gi RWO longhorn-paperless

namespace: rabbitmq

PVC Größe Mode StorageClass
rabbitmq-data-rabbitmq-0 5 Gi RWO longhorn

namespace: zitadel

PVC Größe Mode StorageClass
postgres-data-postgres-0 10 Gi RWO longhorn

Longhorn Pods nach Node

efsckubnode1 (10.42.71.51)

Pod Typ Replicas
longhorn-manager DaemonSet 1
longhorn-csi-plugin DaemonSet 1
engine-image DaemonSet 1
instance-manager – 1
longhorn-driver-deployer Deployment 1
longhorn-ui Deployment 2
csi-attacher Deployment 3
csi-provisioner Deployment 3
csi-resizer Deployment 3
csi-snapshotter Deployment 3

efsckubnode2 (10.42.71.52)

Pod Typ Beschreibung
longhorn-manager DaemonSet –
longhorn-csi-plugin DaemonSet –
engine-image DaemonSet –
instance-manager – –
share-manager (erpnext) – Bedient RWX-Volume erpnext (3Gi)
share-manager (erpnext-logs) – Bedient RWX-Volume erpnext-logs (1Gi)
daily-backup CronJob Tägliches Backup

share-manager Pods werden für RWX-Volumes benötigt: Longhorn stellt RWX über NFS-Share-Manager bereit.
CSI-Controller-Pods (attacher, provisioner, resizer, snapshotter) laufen nur auf efsckubnode1.

Hinweise

  • Longhorn UI: erreichbar über Port-Forward kubectl port-forward -n longhorn-system svc/longhorn-frontend 8080:80
  • Replikation: Standard 2 Replicas (beide Worker-Nodes)
  • Backups: daily-backup CronJob auf efsckubnode2
Reference in New Issue View Git Blame Copy Permalink