docs(esh-pve): GPU swap done — T400 out, RTX 2000 Ada in; VM 102 leftovers cleared after the reboot

This commit is contained in:
vh
2026-09-24 21:38:36 -07:00
parent 719e3fa324
commit 7cbd4c3781
+13 -11
View File
@@ -38,17 +38,19 @@ could not remove (`pve/vm-102-disk-1` 256G, `pve/vm-102-disk-2` TPM). `qm status
102` said `stopped` the whole time. **Only a host reboot clears it.** It is the
same class of fault as the August freeze: VFIO plus this T400.
**Reboot planned for the evening of 2026-09-24 with Prime on site.** Checklist:
1. Cluster: rebooting makes esh-pve-nas's `/etc/pve` read-only until this node is
back (guests keep running). Accept that, or run `pvecm expected 1` there first.
2. After boot, `lvremove pve/vm-102-disk-1 pve/vm-102-disk-2`, then confirm
~16 GB RAM is freed and nothing holds `01:00.0`. The T400 stays bound to
vfio-pci, unassigned.
3. esh-docker-vm: watch the known NFS boot race and the traefik restart and
crowdsec fail-closed recovery (auto-memory incident notes). DNS for
esh-userland comes back with this host.
4. If Prime cables an onboard RJ45 (the **I226-LM** at `59:00.0` is the AMT port),
provision AMT in MEBx during the same window. See the vPro section.
**Done 2026-09-24 2119–2137 (Prime on site).** Guests were shut down cleanly and
the host was powered off. The T400 was **removed** and replaced by an **NVIDIA RTX
2000 / 2000E Ada** (`01:00.0`, `10de:28b0`, subsystem `10de:1871`). The host was
back at 2136:49. The reboot cleared the hung `kvm -id 102`, and the two leftover
LVs were then removed; host RAM in use fell to ~4 GB before esh-vm-docker was
fully up. esh-vm-docker, esh-vm-db and esh-scale all started by themselves.
esh-vm-docker's NFS (`/mnt/books`, `/mnt/backup`) mounted, traefik served with no
403s, and AdGuard resolved.
The new card has **no driver bound**. `/etc/modprobe.d` still lists the T400's IDs
for vfio-pci (`10de:1ff2,10de:10fa`), which no longer match anything; harmless, but
clean it up when the card's use is decided. The intended use is a local
embedding endpoint.
## Watchdog — hardware, not software