feat(esh-pve-2): drop stale 1 TB boot entry, wipe it into LVM-thin storage vmstore

2026-10-02 22:08:19 -07:00
parent 6315ce38d5
commit e707d87713
1 changed files with 85 additions and 0 deletions
@@ -0,0 +1,85 @@
+# esh-pve-2 (MS-03, ESH) disk prep, 2026-10-02, at Prime's request:
+#   - drop the firmware boot entry for the old Proxmox install on the 1 TB drive;
+#   - wipe that drive and turn it into LVM-thin VM storage `vmstore`.
+#   scripts/elway infra-ops@10.0.10.70 --playbook playbooks/esh-pve-2-disk-prep.yaml
+#
+# Disks (read live before writing this; nvme0/nvme1 numbering can swap between boots, so every
+# destructive step names the drive by its by-id link, which carries the serial):
+#   boot  KXG60ZNV256G NVMe TOSHIBA 256GB  serial 199A3537K01N  ESP PARTUUID a64d873f-…  VG pve (root)
+#   data  CT1000E100SSD8 (Crucial 1 TB)    serial 2611EAD0291A  ESP PARTUUID 56c18061-…  VG pve-OLD-2E68F512
+# The old VG is an earlier PVE install the installer renamed; its thin pool was 0.00% used (no guest disks).
+# The GRUB menu itself never listed the old install (os-prober is not installed and is disabled), so
+# the only stale entry is the firmware's Boot0005 "UEFI OS" pointing at the 1 TB drive's ESP.
+#
+# Re-runnable: once `vmstore` exists the wipe steps skip, so a second run cannot touch the new pool.
+steps:
+  - name: refuse unless the box booted from, and has root and /boot/efi on, the 256 GB drive
+    sudo: true
+    shell: |
+      set -e
+      efibootmgr | grep -q '^BootCurrent: 0000'
+      efibootmgr -v | grep -q '^Boot0000\* proxmox.*a64d873f-e2d7-4da6-882c-cef6767fb77f'
+      test "$(findmnt -no PARTUUID /boot/efi)" = a64d873f-e2d7-4da6-882c-cef6767fb77f
+      lsblk -nso SERIAL /dev/mapper/pve-root | grep -q 199A3537K01N
+      ! lsblk -nso SERIAL /dev/mapper/pve-root | grep -q 2611EAD0291A
+      ! lsblk -nro MOUNTPOINTS /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A | grep -q .
+    changed_when: "false"
+
+  - name: delete firmware entry Boot0005 (old install's fallback loader on the 1 TB drive)
+    sudo: true
+    shell: efibootmgr -b 0005 -B
+    when: "efibootmgr -v | grep -q '^Boot0005.*56c18061-9c09-427f-8862-f531a558f630'"
+
+  - name: deactivate and remove the old install's volume group
+    sudo: true
+    shell: |
+      set -e
+      vgchange -an pve-OLD-2E68F512
+      vgremove -y pve-OLD-2E68F512
+    when: "vgs pve-OLD-2E68F512 >/dev/null 2>&1"
+
+  - name: remove the LVM label from the old PV
+    sudo: true
+    shell: pvremove -y /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part3
+    when: "pvs /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part3 >/dev/null 2>&1"
+
+  - name: wipe the 1 TB drive (partition signatures, GPT, then a full discard)
+    sudo: true
+    shell: |
+      set -e
+      wipefs -a /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part1 /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part2 /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part3
+      sgdisk --zap-all /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A
+      wipefs -a /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A
+      blkdiscard -f /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A
+      blockdev --rereadpt /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A || true   # PVE ships no partprobe
+      udevadm settle
+    when: "! vgs vmstore >/dev/null 2>&1 && test -e /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A-part3"
+
+  - name: create LVM-thin pool + storage `vmstore` (the GUI's Disks > LVM-Thin > Create path)
+    sudo: true
+    shell: pvesh create /nodes/esh-pve-2/disks/lvmthin --device /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A --name vmstore --add_storage 1
+    when: "! vgs vmstore >/dev/null 2>&1"
+
+verify:
+  - name: no firmware entry points at the old ESP; Boot0000 (256 GB) is still first
+    sudo: true
+    shell: |
+      ! efibootmgr -v | grep -q 56c18061-9c09-427f-8862-f531a558f630 && efibootmgr | grep -q '^BootOrder: 0000'
+    changed_when: "false"
+  - name: old VG gone; vmstore VG lives on the 1 TB drive
+    sudo: true
+    shell: |
+      ! vgs pve-OLD-2E68F512 >/dev/null 2>&1 && lsblk -nro NAME /dev/disk/by-id/nvme-CT1000E100SSD8_2611EAD0291A | grep -q '^vmstore-vmstore$'
+    changed_when: "false"
+  - name: vmstore is an active storage with images,rootdir content
+    sudo: true
+    shell: |
+      pvesm status --storage vmstore | grep -q ' active ' && grep -A4 '^lvmthin: vmstore' /etc/pve/storage.cfg | grep -q 'content rootdir,images\|content images,rootdir'
+    changed_when: "false"
+  - name: positive control — allocate and free a 1 GiB volume on vmstore
+    sudo: true
+    shell: |
+      set -e
+      pvesm alloc vmstore 9999 vm-9999-disk-0 1G
+      pvesm free vmstore:vm-9999-disk-0
+    changed_when: "false"