fix: correct storage bus types after deep-dive on tsys6/7
Verified via lsblk TRAN column and dmesg that tsys6/7 local-lvm is
backed by WD My Passport portable HDDs on USB 2.0 (~30 MB/s), not
internal SATA/SAS as previously assumed. Both R610 and R620 only have
EHCI controllers -- no USB 3.0/xHCI. tsys7 has 5 empty SATA ports but
no drives installed; tsys6's SAS controller is disabled in BIOS.
Corrected all wnode recommendations: tsys6/7 stay on NFS (faster than
their USB 2.0 local-lvm). Only tsys3/5/9/2 have local storage suitable
for VM boot disks.
Also discovered tsys3 boots from Samsung PM961 NVMe 512GB (previously
documented as HDD). Its 349 GB local-lvm is the fastest existing local
storage in the fleet.
Updated speed tier table, wnode placement tables, storage distribution
matrices, migration plan, and caching strategy across PROJECT.md and
K8S.md.
💘 Generated with Crush
Assisted-by: Crush:glm-5.2
This commit is contained in:
@@ -59,11 +59,14 @@ on a VPS in Reston VA (Cloudron) -- this cluster is R&D only.
|
||||
if tsys4 goes down. Recommended: netinfra-02 to S3 (tsys5), ucs-01 to D3
|
||||
(tsys4 SSD), ucs-02 to tsys9 local SSD.
|
||||
3. **No k8s node uses SSD or NVMe.** All on NFS-over-HDD. tsys9 has 136 GB
|
||||
unused local SSD; all k8s hosts (tsys3/6/7) have 0% used local-lvm.
|
||||
unused local SSD; tsys3 has 349 GB unused local **NVMe** (Samsung PM961).
|
||||
4. **D3 SSD (tsys4) is 99% empty (445 GB free).** Nobody using the only SSD
|
||||
NFS export. Should host ucs-01 (LDAP/AD latency benefit).
|
||||
5. **S2/S3/S4 (tsys5) are all 99% empty.** Secondary storage server barely
|
||||
used. VMs should be rebalanced to reduce tsys4 blast radius.
|
||||
6. **tsys6/7 local-lvm is USB 2.0 portable HDD** (WD My Passport, ~30 MB/s).
|
||||
Slower than NFS-HDD. Do NOT use for VM storage. Entire Proxmox OS on these
|
||||
hosts boots from USB 2.0 -- user accepts this risk.
|
||||
|
||||
## Storage network IPs (VLAN1000, 10.100.100.0/24)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user