Installation Manual - 22 Spoke ODF RAID0 Disk Prep

How to prepare spoke-dc-v7 physical worker disks for ODF by creating one RAID0 volume per intended data disk.

This chapter prepares the spoke-dc-v7 physical worker disks for ODF.

The previous validation found that each physical worker exposed one 4.8 TB RAID5 logical volume from the HPE MR408i-o Gen11 controller. For this deployment, the chosen layout is one RAID0 logical volume per intended data disk, so RHCOS sees three separate 2.4 TB devices on each physical worker.

Do not touch the HPE NS204i-u Gen11 Boot Controller. It contains the NVMe boot volume.

Target State

Expected RHCOS view on each physical worker:

• nvme0n1: OS boot volume, mounted;
• sda, sdb, sdc: 2.4 TB MR408i-o Gen11 rotational data devices, unmounted and unformatted.

Important Lessons

• Redfish RAIDType=None was advertised but was not usable on this controller in this environment.
• One-disk RAID0 volume creation worked and produced the desired separate host-visible data devices.
• After Storage.ResetToDefaults, the controller may need a reboot before volume creation works reliably.
• After reset, drive IDs can change. Re-read the controller drive list before creating volumes.
• Create and validate the layout one worker at a time.

Prerequisites

• HTPasswd admin access works.
• The cluster is healthy before starting.
• BMC credentials are available from local secret custody or Vault.
• No ODF data exists yet.
• You have identified the intended three data disks per worker.

Create a temporary admin kubeconfig:

tmp="$(mktemp)"
read -rsp "HTPasswd password for ze: " ZE_HTPASSWD_PASSWORD
printf '\n'

KUBECONFIG="$tmp" oc login \
  https://api.spoke-dc-v7.v7.comptech-lab.com:6443 \
  -u ze \
  -p "$ZE_HTPASSWD_PASSWORD" \
  --insecure-skip-tls-verify=true

export KUBECONFIG="$tmp"
unset ZE_HTPASSWD_PASSWORD

Validate cluster health:

oc get nodes
oc get co | awk 'NR==1 || $3!="True" || $4!="False" || $5!="False"'
oc get mcp

Worker Procedure

Run the procedure one worker at a time.

Map:

After reset, confirm the drive IDs before creating RAID0 volumes.

Cordon And Drain

oc adm cordon "$NODE"
oc adm drain "$NODE" \
  --ignore-daemonsets \
  --delete-emptydir-data \
  --force \
  --timeout=10m

Reset Only The MR Data Controller

Do not run this against the DE009000 NS204i boot controller.

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  -H 'Content-Type: application/json' \
  -X POST \
  -d '{"ResetType":"ResetAll"}' \
  "https://${BMC_HOST}/redfish/v1/Systems/1/Storage/${MR_STORAGE_ID}/Actions/Storage.ResetToDefaults"

Poll the task until complete:

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  "https://${BMC_HOST}/redfish/v1/TaskService/Tasks/1" \
  | jq '{TaskState, TaskStatus, EndTime}'

Reboot To Settle The Controller

If the volume collection returns intermittent 405, 503, or internal errors, reboot the server while it is still cordoned and drained.

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  -H 'Content-Type: application/json' \
  -X POST \
  -d '{"ResetType":"GracefulRestart"}' \
  "https://${BMC_HOST}/redfish/v1/Systems/1/Actions/ComputerSystem.Reset"

Wait until the node returns Ready,SchedulingDisabled.

oc get node "$NODE" -w

Re-Read Drive IDs

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  "https://${BMC_HOST}/redfish/v1/Systems/1/Storage/${MR_STORAGE_ID}" \
  | jq -r '.Drives[]?."@odata.id"'

Inspect each intended drive:

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  "https://${BMC_HOST}/redfish/v1/Chassis/${MR_STORAGE_ID}/Drives/${DRIVE_ID}" \
  | jq '{Id, Name, CapacityBytes, Status, PhysicalLocation}'

Create One RAID0 Volume Per Data Disk

for drive in "${DRIVES[@]}"; do
  uri="/redfish/v1/Chassis/${MR_STORAGE_ID}/Drives/${drive}"

  jq -n \
    --arg name "RAID0-DATA-${drive}" \
    --arg uri "$uri" \
    '{Name:$name, RAIDType:"RAID0", Links:{Drives:[{"@odata.id":$uri}]}}' \
    > /tmp/raid0-volume.json

  curl -sk -u "${BMC_USER}:${BMC_PASS}" \
    -H 'Content-Type: application/json' \
    -X POST \
    -d @/tmp/raid0-volume.json \
    "https://${BMC_HOST}/redfish/v1/Systems/1/Storage/${MR_STORAGE_ID}/Volumes"
done

Expected Redfish result:

• three volumes;
• each volume has RAIDType=RAID0;
• each volume is backed by exactly one drive;
• each volume has Status.State=Enabled and Status.Health=OK.

Reboot And Validate RHCOS View

curl -sk -u "${BMC_USER}:${BMC_PASS}" \
  -H 'Content-Type: application/json' \
  -X POST \
  -d '{"ResetType":"GracefulRestart"}' \
  "https://${BMC_HOST}/redfish/v1/Systems/1/Actions/ComputerSystem.Reset"

Wait until the node returns Ready,SchedulingDisabled, then validate devices:

oc debug --quiet "node/${NODE}" -- chroot /host \
  lsblk -b -o NAME,TYPE,SIZE,MODEL,SERIAL,ROTA,TRAN,WWN,PKNAME,MOUNTPOINTS,FSTYPE

Expected:

• sda, sdb, sdc exist;
• each is approximately 2399913639936 bytes;
• each has model MR408i-o Gen11;
• none is mounted;
• the NVMe boot volume remains mounted as the OS disk.

Uncordon

oc adm uncordon "$NODE"
oc get node "$NODE"

Recorded Result

The recorded spoke-dc-v7 result:

Validation report:

reports/spoke-dc-v7-odf-raid0-disk-prep-20260516.md

Cleanup

Remove temporary kubeconfigs and JSON payloads:

rm -f "$tmp" /tmp/raid0-volume.json
unset KUBECONFIG

Completion State

For the recorded spoke-dc-v7 run:

• issue: OP-GF-SPOKEDCV7-9;
• storage decision: one RAID0 logical volume per intended ODF data disk;
• all physical workers expose three separate 2.4 TB data devices;
• all physical workers returned Ready;
• ODF disk preparation is complete for initial LSO/ODF planning.