KNEL/PFVCluster
5
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs(goals): codify local storage policy — ephemeral scratch, not OS drives

Browse Source 
Add CAPACITY-GOALS.md §6: local storage is a scratch tier for ephemeral workloads
(container image cache, build artifacts, emptyDir), NOT for VM OS drives (which
stay on NFS where redundancy is handled via 01/02 pairs + cross-disk spreading).
Documents per-host local storage suitability: tsys3/9/2 excellent (NVMe/SSD),
tsys5 usable (HDD bulk), tsys6/7 unusable (USB 2.0 at 30 MB/s). Includes the
future k8s StorageClass mapping (local-fast for scratch, nfs-hdd for persistent).
Adds forward-looking §12 to the audit referencing this policy.

💘 Generated with Crush

Assisted-by: Crush:glm-5.2
This commit is contained in:
mrcharles
2026-07-29 16:33:56 -05:00
parent edb34a75ec
commit 9c47d93a2f
2 changed files with 122 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
proxmox/docs/AUDIT-2026-07-29.md
+32
View File
@@ -418,3 +418,35 @@ HA pairs stay split:
- `kubectl top nodes` — confirm cnodes healthy
- `ssh root@pfv-tsys4 'df -h /mnt/tsys4/D2'` — confirm D2 frees up
- `ssh root@pfv-tsys5 'df -h /mnt/pfv-tsys5/S2 /mnt/pfv-tsys5/S3 /mnt/pfv-tsys5/S4'` — confirm fills
---
## 12. Future: local storage as ephemeral scratch
> Codified as standing policy in [`CAPACITY-GOALS.md`](CAPACITY-GOALS.md) §6.
The fleet has substantial **unused local storage IOPS** on several hosts. OS
drives stay on NFS (no per-node disk redundancy — see policy), but local storage
is ideal for **ephemeral scratch** (container image cache, build artifacts,
`emptyDir` volumes) where data loss on node failure is acceptable.
**Per-host opportunity:**
| Host | Local | Available | Scratch-suitable? |
|------|-------|-----------|-------------------|
| tsys3 | NVMe (PM961) | 349 GB | **Yes — best existing** |
| tsys9 | SSD (PNY CS900) | 136 GB | **Yes** |
| tsys2 | NVMe + SSD | 512 GB + 1 TB | **Yes (incoming)** |
| tsys5 | HDD (Hitachi) | ~1.7 TB | Yes (bulk) |
| tsys6 | USB 2.0 HDD | 1.7 TB | **NO (30 MB/s)** |
| tsys7 | USB 2.0 HDD | 1.7 TB | **NO (30 MB/s)** |
**The tsys6/7 problem:** they are the RackRental hosts (most in need of fast
image pulls) but their local storage is USB 2.0 — slower than NFS. The real fix
is installing internal drives (tsys6 has a disabled SAS controller; tsys7 has 5
empty SATA ports). Deferred to a physical-hardware session.
**Implementation:** post-Friday, once SSD/NVMe tiers are installed and the k8s
StorageClass design is finalized. Wnodes with local storage get a `/scratch`
mount for ephemeral data; persistent data stays on NFS. See
[`CAPACITY-GOALS.md`](CAPACITY-GOALS.md) §6.4 for the StorageClass mapping.