OPTIONAL: This procedure is only applicable if Fabric Manager nodes were deployed during the CSM installation.
Although Fabric Manager Nodes (FMNs) were deployed during the CSM installation, the initial deployment would not include the final image with all necessary components. Once the other HPE Cray EX software products have been installed via the Install and Upgrade Framework (IUF), the FMNs need to be redeployed with the new customized image.
Follow the procedure in the FMN Base Image Creation section of Configure FM (Fabric Manager) On Baremetal to build the new FMN base image.
This procedure will:
sat bootprep configuration file for FMNsat bootprep run to generate the new FMN image and upload it to S3Once the new FMN image has been built and uploaded to S3, update the boot parameters in the Boot Script Service (BSS) to point the FMNs to the new image.
(ncn-mw#) Set an environment variable for the new IMS image ID.
After running sat bootprep run, obtain the IMS resultant image ID from the output or from the CFS session:
NEW_IMS_IMAGE_ID="<IMS_IMAGE_ID_FROM_BOOTPREP>"
(ncn-mw#) Determine the component names (xnames) of the FMNs.
cray hsm state components list --role Management --subrole FabricManager --format json | jq -r '.Components[].ID'
Example output:
x3000c0s28b0n0
x3000c0s29b0n0
(ncn-mw#) Update the boot parameters for the FMNs.
Replace the <xname> placeholders with the actual xnames of the FMNs.
/usr/share/doc/csm/scripts/operations/node_management/assign-ncn-images.sh \
-p "${NEW_IMS_IMAGE_ID}" <xname1> <xname2>
For example:
/usr/share/doc/csm/scripts/operations/node_management/assign-ncn-images.sh \
-p "${NEW_IMS_IMAGE_ID}" x3000c0s28b0n0 x3000c0s29b0n0
(ncn-mw#) Verify the boot parameters have been updated.
cray bss bootparameters list --name <xname> --format json | jq -r '.[0].params' | grep metal.server
The output should show the new IMS image ID in the metal.server parameter.
(ncn-mw#) Set metal.no-wipe=0 to allow the disk to be wiped during redeployment.
For each FMN, set metal.no-wipe=0:
For Example:
TARGET_XNAME=x3000c0s28b0n0
csi handoff bss-update-param --set metal.no-wipe=0 --limit ${TARGET_XNAME}
Expected output:
2026/06/05 11:33:35 TOKEN was not set. Attempting to read API token from Kubernetes directly ...
2026/06/05 11:33:35 Getting management NCNs from SLS...
12
2026/06/05 11:33:35 Done getting management NCNs from SLS.
2026/06/05 11:33:35 Updating NCN kernel parameters...
2026/06/05 11:33:35 Successfully PUT BSS entry for x3000c0s28b0n0
2026/06/05 11:33:35 Done updating NCN kernel parameters.
Repeat for each FMN.
(ncn-mw#) Verify the change:
cray bss bootparameters list --name ${TARGET_XNAME} --format=json | jq -r '.[0].params' | grep metal.no-wipe
The output should show metal.no-wipe=0.
After updating BSS with the new image and setting metal.no-wipe=0, redeploy the FMNs to apply the new image.
Follow the Boot NCN procedure for each Fabric Manager node. This procedure will:
metal.no-wipe=1 after successful boot to preserve data on future rebootsNote: Skip the sections in Boot NCN that are specific to master, worker, or storage nodes (such as verifying cluster membership or Ceph operations).
After the Fabric Manager nodes have been redeployed and are running with the new image, join them to Spire to avoid issues with Spire tokens.
(ncn-mw#) Join Spire on the Fabric Manager nodes.
/opt/cray/platform-utils/spire/fix-spire-on-fmn.sh