This section describes how to repurpose a compute node to be used as a User Access Node (UAN). This is typically done when the processor type of the compute node is not yet available in a UAN server.
The following steps outline the process of repurposing a compute node to be used as a UAN.
Verify the System Default Route is set to CHN
.
Change the role of the compute node in the Hardware State Manager from Compute
to Application
and set the sub-role to UAN
.
Ensure that IPs on the CHN exist for the computes nodes in SLS.
Boot the repurposed compute node as a UAN.
Verify the repurposed compute node functions as a UAN.
There are no changes needed in hardware, network cabling, or UEFI/BIOS/BMC configuration to repurpose a compute node for use as a UAN. However, compute nodes do not have the necessary network interface cards to support user access over the Customer Access Network (CAN). Additionally, the network configuration of Mountain Cabinets do not support the CAN network. Therefore, repurposing a compute node as a UAN requires the system to be configured to use the Customer High-Speed Network (CHN) and that the compute nodes have a CHN IP address in SLS.
SystemDefaultRoute
must be CHN
uan_can_setup
must be set to true
in the uan-config-management repoPerform the following steps to repurpose a compute node for use as a UAN.
Log in to the master node ncn-m001
. All commands in this procedure are run from the master node.
Verify the system is configured to use the CHN
as the System Default Route. If the SystemDefaultRoute
is not CHN
, the compute nodes may not be repurposed as UAN.
ncn-m001# cray sls networks describe BICAN --format json | jq -r '.ExtraProperties.SystemDefaultRoute'
Verify a CHN IP address exists in SLS for each repurposed compute node. Repeat the following command and replace <XNAME>
with the xname of each repurposed compute node. The compute node must have a CHN IP address in SLS or it cannot be repurposed as a UAN. See Add Compute IP addresses to CHN SLS data
section of the Cray System Management documentation for information on adding compute nodes to the CHN.
ncn-m001# cray sls networks describe CHN | q -r '.ExtraProperties.Subnets[] | select(.FullName == "CHN Bootstrap DHCP Subnet") | .IPReservations[] | select(.Comment == "<XNAME>")'
Verify that uan_can_setup: true
is set in the uan-config-management
CFS repo. See Enabling the Customer Access Network (CAN) or the Customer High Speed Network (CHN) for more information.
Change the role and sub-role in HSM of the compute node(s) being repurposed as UANs to Application
and UAN
, respectively. Repeat the following command and replace <XNAME>
with the xname of each repurposed compute node.
ncn-m001# cray hsm state components role update --role Application --sub-role UAN <XNAME>
Verify the role and sub-role in HSM of the repurposed compute node(s) has been changed to ‘Application and 'UAN
, respectively. Repeat the following command and replace <XNAME>
with the xname of each repurposed compute node.
ncn-m001# cray hsm state components describe <XNAME>
Run the BOS session template used to boot the UAN nodes. See Boot UAN Nodes for more information on booting UAN nodes with BOS. Replace <UAN_SESSIONTEMPLATE>
with the name of the BOS session template used to boot the UAN nodes and <XNAME>
with the xname of the repurposed compute node.
ncn-m001# cray bos session create --template-uuid <UAN_SESSIONTEMPLATE> --operation reboot --limit <XNAME>
Once the repurposed compute node is booted as a UAN, the following steps will verify it is configured as a UAN. These steps may vary dependent upon how the site has configured the UAN nodes.
Verify the repurposed compute node has finished the configuration phase. The output should “configured”.
ncn-m001# cray cfs components describe <XNAME> --format json | jq -r .configurationStatus
Login to the repurposed compute node from the master node ncn-m001
as the root user.
Verify that the hsn0
interface has the CHN IP address assigned to it in SLS.
uan# ip a | grep hsn0
Verify the default route is via hsn0
uan# ip r | grep default
Verify that all site UAN filesystems are mounted.
If LDAP is used for user authentication, verify the LDAP service is reachable.
uan# ping <ldap_service_ip>
If SLURM is used, test sinfo
and srun
commands. This example srun
command should return the hostname of 4 compute nodes.
uan# sinfo
uan# srun -N4 hostname
Login to the repurposed compute node as an authorized non-root user from any host that should have UAN access.
If SLURM is used, test sinfo
and srun
commands. This example srun
command should return the hostname of 4 compute nodes.
uan# sinfo
uan# srun -N4 hostname