Cray System Management Documentation > Cray System Management (CSM) Administration Guide > node management > Add Remove Replace NCNs > Add NCN Data

Add NCN Data

Topics

Description

Add NCN data to the System Layout Service (SLS), Boot Script Service (BSS), and Hardware State Manager (HSM) as needed, in order to add an NCN to the system.

Scenarios where this procedure is applicable:

Adding a management NCN that has not previously been in the system:
- Add an additional NCN to an existing cabinet
- Add an NCN that is replacing another NCN of the same type and in the same slot
- Add a new NCN that replaces an NCN removed from the system in a different location
Adding a management NCN that has been present in the system previously:
- Add an NCN that was previously removed from the system to move it to a new location

Prerequisites

The Cray command line interface (CLI) tool is initialized and configured on the system. See Configure the Cray CLI.
The latest CSM documentation is installed on the system. See Check for latest documentation.

Procedure

(ncn-mw#) Retrieve an API token:

export TOKEN=$(curl -s -S -d grant_type=client_credentials \
                -d client_id=admin-client -d client_secret=`kubectl get secrets admin-client-auth \
                -o jsonpath='{.data.client-secret}' | base64 -d` \
                https://api-gw-service-nmn.local/keycloak/realms/shasta/protocol/openid-connect/token \
                | jq -r '.access_token')

Collect information from the NCN

(ncn-mw#) Determine the component name (xname) of the NCN, if it has not been determined yet, by referring to the HMN tab of the system’s SHCD spreadsheet.

Sample row from the HMN tab of an SHCD file:

Source (J20)	Source Rack (K20)	Source Location (L20)	(M20)	Parent (N20)	(O20)	Source Port (P20)	Destination (Q20)	Destination Rack (R20)	Destination Location (S20)	(T20)	Destination Port (U20)
`wn01`	`x3000`	`u04`	`-`			`j3`	`sw-smn01`	`x3000`	`u14`	`-`	`j48`

The Source name for a worker NCN would be in the format wn01; master NCNs have format mn01 and storage NCNs have format sn01.

Node xname format: xXcCsSbBnN

		SHCD Column to Reference	Description
X	Cabinet number	Source Rack (K20)	The Cabinet or rack number containing the Management NCN.
C	Chassis number		For air-cooled nodes within a standard rack, the chassis is `0`. If the air-cooled node is within an air-cooled chassis in an
S	Slot/Rack U	Source Location (L20)	The Slot of the node is determined by the bottom most rack U that node occupies.
B	BMC number		For Management NCNs the BMC number is 0.
N	Node number		For Management NCNs the Node number is 0.

XNAME=x3000c0s4b0n0

(ncn-mw#) Skip if adding ncn-m001: Determine the NCN BMC xname by removing the trailing n0 from the NCN xname:
```
BMC_XNAME=x3000c0s4b0
```

(ncn-mw#) Skip if adding ncn-m001: Determine the xname of the MgmtSwitchConnector (the switch port of the management switch that the BMC is connected to). This is not required for ncn-m001, because its BMC is typically connected to the site network.

Sample row from the HMN tab of an SHCD:

Source (J20)	Source Rack (K20)	Source Location (L20)	(M20)	Parent (N20)	(O20)	Source Port (P20)	Destination (Q20)	Destination Rack (R20)	Destination Location (S20)	(T20)	Destination Port (U20)
`wn01`	`x3000`	`u04`	`-`			`j3`	`sw-smn01`	`x3000`	`u14`	`-`	`j48`

MgmtSwitchConnector xname format: xXcCwWjJ

		SHCD Column to Reference	Description
X	Cabinet number	Destination Rack (R20)	The Cabinet or rack number containing the management NCN.
C	Chassis number		For air-cooled management switches within standard racks, the chassis is `0`. If the air-cooled management switch is within an air-cooled chassis
W	Slot/Rack U	Destination Location (S20)	The Slot/Rack U that the management switch occupies.
J	Switch port number	Destination Port (U20)	The switch port on the switch that the NCN BMC is cabled to.

MGMT_SWITCH_CONNECTOR=x3000c0w14j48

(ncn-mw#) Skip if adding ncn-m001: Determine the xname of the management switch by removing the trailing jJ from the MgmtSwitchConnector xname.
```
MGMT_SWITCH=x3000c0w14
```
(ncn-mw#) Skip if adding ncn-m001: Collect the BMC MAC address.
NOTES
- If the NCN was previously in the system, recall the BMC MAC address recorded from the Remove NCN Data procedure.
- Alternatively, view the MAC address table on the management switch that the BMC is cabled to.
1. (ncn-mw#) Determine the alias of the management switch that is connected to the BMC.
```
cray sls hardware describe "${MGMT_SWITCH}" --format json | jq .ExtraProperties.Aliases[] -r
```
  Example output:
```
sw-leaf-bmc-001
```
2. (ncn-mw#) SSH into the management switch that is connected to the BMC.
```
ssh admin@sw-leaf-bmc-001.hmn
```
3. (sw-leaf-bmc#) Locate the switch port that the BMC is connected to and record its MAC address.
  
  In the commands below, change the value of 1/1/39 to match the BMC’s actual switch port number (the BMC switch port number is the J value in the in the MgmtSwitchConnector xname xXwWjJ). For example: A MgmtSwitchConnector xname of x3000c0w14j39, the switch port number would be 39. In this case,1/1/39 would be used instead of 1/1/48 in the following commands (j39=1/1/39).
  - Dell Management Switch
```
show mac address-table | grep 1/1/48
```
    Example output:
```
4    a4:bf:01:65:68:54    dynamic        1/1/48
```
  - Aruba Management Switch
```
show mac-address-table | include 1/1/48
```
    Example output:
```
a4:bf:01:65:68:54    4        dynamic                   1/1/48
```
(ncn-mw#) Skip if adding ncn-m001: Set the BMC_MAC environment variable to the BMC MAC address.
```
BMC_MAC=a4:bf:01:65:68:54
```

(ncn-mw#) Skip if adding ncn-m001: Determine the current IP address of the NCN BMC.

Query Kea for the BMC MAC address to determine its current IP address.

BMC_IP=$(curl -sk -H "Authorization: Bearer ${TOKEN}" -X POST -H "Content-Type: application/json" \
             -d '{ "command": "lease4-get-all", "service": [ "dhcp4" ] }' \
             https://api-gw-service-nmn.local/apis/dhcp-kea |
             jq --arg BMC_MAC "${BMC_MAC}" \
                 '.[].arguments.leases[] | select(."hw-address" == $BMC_MAC)."ip-address"' -r)
echo ${BMC_IP}

Example output:

10.254.1.26

Troubleshooting If the MAC addresses of the BMC are not present in Kea, then check for the following items:

Verify that the BMC is powered up and has an active connection to the network.

Verify that the BMC is set to DHCP instead of a static IP address.

(ncn-mw#) Ping the BMC to see if it is reachable.
```
ping "${BMC_IP}"
```

(ncn-mw#) Perform this step if adding ncn-m001, otherwise skip: Set the BMC_IP environment variable to the current IP address or hostname of the BMC. This is not the allocated HMN address for the BMC of ncn-m001.
```
BMC_IP=10.0.0.10
```
(ncn-mw#) Perform this step if adding ncn-m001, otherwise skip: At this point the BMC is reachable at a known IP addresses. If adding an NCN that was not previously in the system, then follow the Access and Update the Settings for Replacement NCNs procedure.

Saving/Reloading

(ncn-mw#) Save the currently collected values to a loadable script for reloading.

mkdir -pv /metal/recovery/scratch
cat << EOF > /metal/recovery/scratch/${NODE}-env-vars.sh
#!/bin/bash
export BMC_XNAME=${BMC_XNAME}
export BMC_MAC=${BMC_MAC}
export NODE=${NODE}
export MGMT_SWITCH_CONNECTOR=${MGMT_SWITCH_CONNECTOR}
export XNAME=${XNAME}
EOF

(ncn-mw#) if resuming this procedure, reload the values.

NODE=ncn-w004
. "/metal/recovery/scratch/${NODE}-env-vars.sh"

Collect MAC addresses from the NCN

The collected MAC addresses will be used with the add_management_ncn.py script.

Depending on the hardware present in the NCN, not all of these interfaces may be present. See NCN networking for more information.

NCNs will have either two or four management network interfaces split between two PCIe cards for redundancy, and will utilize only two of their available ports (port 1 of each PCIe card).
Kubernetes Worker NCNs will have high-speed network (HSN) interfaces in addition to their management network interfaces.
Storage NCNs have bond1 members (sun0 and sun1), but only if they have four NICs (port 2 of each PCIe card).

NOTE In rare cases, where a PCIe redundancy goal could not be met, an NCN may be observed using two ports off a single PCIe card. The plan of record design (as detailed in the plan of record page) is to have PCIe redundancy in the event that an entire PCIe card fails, the network will still stay up until maintenance can be performed properly. In the event that a single PCIe card is used, use the first table below as a reference.

NCN with a single PCIe card (two cards with one port each for two ports total):

Interface	CLI Flag	Required MAC Address	Description
`mgmt0`	`--mac-mgmt0`	Required	First MAC address for the MGMT network bond (`bond0`).
`mgmt1`	`--mac-mgmt1`	Required	Second MAC address for the MGMT network bond (`bond0`).
`hsn0`	`--mac-hsn0`	Required for Worker NCNs	MAC address of the first HSN NIC.
`hsn1`	`--mac-hsn1`	Optional for Worker NCNs	MAC address of the second HSN NIC.
`lan0`	`--mac-lan0`	Optional	MAC address for the first site-network.
`lan1`	`--mac-lan1`	Optional	MAC address for the second site-network.
`lan2`	`--mac-lan2`	Optional	MAC address for the third site-network.
`lan3`	`--mac-lan3`	Optional	MAC address for the forth site-network.

NCN with a dual PCIe cards (two cards with two ports each for four ports total):

Interface	CLI Flag	Required MAC Address	Description
`mgmt0`	`--mac-mgmt0`	Required	First MAC address for the MGMT network bond (`bond0`).
`mgmt1`	`--mac-mgmt1`	Required	Second MAC address for the MGMT network bond (`bond0`).
`sun0`	`--mac-sun0`	Required for Storage NCNs	First MAC address for the SUN network bond (`bond1`).
`sun1`	`--mac-sun1`	Required for Storage NCNs	Second MAC address for the SUN network bond (`bond1`).
`hsn0`	`--mac-hsn0`	Required for Worker NCNs	MAC address of the first High Speed Network NIC.
`hsn1`	`--mac-hsn1`	Optional for Worker NCNs	MAC address of the second High Speed Network NIC.
`lan0`	`--mac-lan0`	Optional	MAC address for the first site-network.
`lan1`	`--mac-lan1`	Optional	MAC address for the second site-network.
`lan2`	`--mac-lan2`	Optional	MAC address for the third site-network.
`lan3`	`--mac-lan3`	Optional	MAC address for the forth site-network.

See swapping/moving an NCN or adding a new NCN depending on the current course of action.

Swapping/Moving an NCN

If the NCN being added is being moved to a new location in the system, then these MAC addresses can be retrieved from backup files generated by the Remove NCN Data procedure.

Recall the previous node xname of the NCN being added:
```
PREVIOUS_XNAME=REPLACE_WITH_OLD_XNAME
```

Retrieve the MAC address for the NCN from the backup files:

cat "/tmp/remove_management_ncn/${PREVIOUS_XNAME}/bss-bootparameters-${PREVIOUS_XNAME}.json" |
    jq .[].params -r | tr " " "\n" | grep ifname

Example output for a worker node with a single management PCIe NIC card:

ifname=hsn0:50:6b:4b:23:9f:7c
ifname=lan1:b8:59:9f:d9:9d:e9
ifname=lan0:b8:59:9f:d9:9d:e8
ifname=mgmt0:a4:bf:01:65:6a:aa
ifname=mgmt1:a4:bf:01:65:6a:ab

Using the example output from above, derive the following CLI flags for a worker NCN:

Interface	MAC Address	CLI Flag
`mgmt0`	`a4:bf:01:65:6a:aa`	`--mac-mgmt0=a4:bf:01:65:6a:aa`
`mgmt1`	`a4:bf:01:65:6a:ab`	`--mac-mgmt1=a4:bf:01:65:6a:ab`
`lan0`	`b8:59:9f:d9:9d:e8`	`--mac-lan0=b8:59:9f:d9:9d:e8`
`lan1`	`b8:59:9f:d9:9d:e9`	`--mac-lan1=b8:59:9f:d9:9d:e9`
`hsn0`	`50:6b:4b:23:9f:7c`	`--mac-hsn0=50:6b:4b:23:9f:7c`

Adding a new NCN

The NCN MAC addresses need to be collected using the Collect NCN MAC Addresses procedure.

Add the NCN to BSS, HSM, and SLS

(ncn-mw#) Perform a dry run of the add_management_ncn.py script in order to determine if any validation failures occur:

NOTE Update the following command with the MAC addresses and interfaces that were collected from the NCN.

For any NCN that does not have a site-link (e.g. ncn-m001):

cd /usr/share/doc/csm/scripts/operations/node_management/Add_Remove_Replace_NCNs/
./add_management_ncn.py ncn-data \
    --xname "${XNAME}" \
    --alias "${NODE}" \
    --bmc-mgmt-switch-connector "${MGMT_SWITCH_CONNECTOR}" \
    --mac-bmc "${BMC_MAC}" \
    --mac-mgmt0 a4:bf:01:65:6a:aa \
    --mac-mgmt1 a4:bf:01:65:6a:ab \
    --mac-hsn0 50:6b:4b:23:9f:7c \
    --mac-lan0 b8:59:9f:d9:9d:e8 \
    --mac-lan1 b8:59:9f:d9:9d:e9

For any NCN with a site-link (e.g. ncn-m001), omit the --bmc-mgmt-switch-connector and --mac-bmc arguments:

cd /usr/share/doc/csm/scripts/operations/node_management/Add_Remove_Replace_NCNs/
./add_management_ncn.py ncn-data \
    --xname "${XNAME}" \
    --alias "${NODE}" \
    --mac-mgmt0 a4:bf:01:65:6a:aa \
    --mac-mgmt1 a4:bf:01:65:6a:ab \
    --mac-lan0 b8:59:9f:d9:9d:e8 \
    --mac-lan1 b8:59:9f:d9:9d:e9

(ncn-mw#) Run the add_management_ncn.py script again, adding the --perform-changes argument to the command run in the previous step:

NOTE Depending on the networking configuration of the system the CMN or CAN networks may not be present in SLS network data. No IP addresses will be allocated for networks that do not exist in SLS.

For example:

./add_management_ncn.py ncn-data \
    --xname "${XNAME}" \
    --alias "${NODE}" \
    --bmc-mgmt-switch-connector "${MGMT_SWITCH_CONNECTOR}" \
    --mac-bmc "${BMC_MAC}" \
    --mac-mgmt0 a4:bf:01:65:6a:aa \
    --mac-mgmt1 a4:bf:01:65:6a:ab \
    --mac-hsn0 50:6b:4b:23:9f:7c \
    --mac-lan0 b8:59:9f:d9:9d:e8 \
    --mac-lan1 b8:59:9f:d9:9d:e9 \
    --perform-changes

Example output:

...
x3000c0s3b0n0 (ncn-m002) has been added to SLS/HSM/BSS
    WARNING The NCN BMC currently has the IP address: 10.254.1.20, and needs to have IP address 10.254.1.13

    =================================
    Management NCN IP Allocation
    =================================
    Network | IP Address
    --------|-----------
    CMN     | 10.103.11.42
    CAN     | 10.102.4.10
    HMN     | 10.254.1.14
    MTL     | 10.1.1.7
    NMN     | 10.252.1.9

    =================================
    Management NCN BMC IP Allocation
    =================================
    Network | IP Address
    --------|-----------
    HMN     | 10.254.1.13

Troubleshooting

If the following text is present at the end of the add_management_ncn.py script output, then the BMC is not using the assigned IP address. This happens because the BMC requests an IP prior to the NCN data being added, and now that we have added data we need to align the BMC.

Sample output when the BMC has an unexpected IP address.
x3000c0s3b0n0 (ncn-m002) has been added to SLS/HSM/BSS
    WARNING The NCN BMC currently has the IP address: <$BMC_IP>, and needs to have IP address X.Y.Z.W
Restart the BMC to pick up the expected IP address:

read -s is used to read the password in order to prevent it from being echoed to the screen or recorded in the shell history.
read -r -s -p "BMC root password: " IPMI_PASSWORD
export IPMI_PASSWORD
ipmitool -U root -I lanplus -E -H "${BMC_IP}" mc reset cold
sleep 60

(ncn-mw#) Skip if adding ncn-m001: Verify that the BMC is reachable at the expected IP address.
```
ping "${NODE}-mgmt"
```
NOTE Wait five minutes for Kea and the HSM to sync. If ping continues to fail, then re-run the previous step to restart the BMC.

(ncn-mw#) Restart the REDS deployment.

kubectl -n services rollout restart deployment cray-reds

Expected output:

deployment.apps/cray-reds restarted

(ncn-mw#) Wait for REDS to restart.

kubectl -n services rollout status  deployment cray-reds

Expected output:

Waiting for deployment "cray-reds" rollout to finish: 1 old replicas are pending termination...
Waiting for deployment "cray-reds" rollout to finish: 1 old replicas are pending termination...
deployment "cray-reds" successfully rolled out

(ncn-mw#) Skip if adding ncn-m001: Wait for the NCN BMC to get discovered by HSM.

NOTE If the BMC of ncn-m001 is connected to the site network, then the BMC will not be discovered, because it is not connected via the HMN network.

watch -n 0.2 "cray hsm inventory redfishEndpoints describe '${BMC_XNAME}' --format json"

Wait until the LastDiscoveryAttempt field is DiscoverOK:

{
    "ID": "x3000c0s38b0",
    "Type": "NodeBMC",
    "Hostname": "",
    "Domain": "",
    "FQDN": "x3000c0s38b0",
    "Enabled": true,
    "UUID": "cc48551e-ec22-4bef-b8a3-bb3261749a0d",
    "User": "root",
    "Password": "",
    "RediscoverOnUpdate": true,
    "DiscoveryInfo": {
        "LastDiscoveryAttempt": "2022-02-28T22:54:08.496898Z",
        "LastDiscoveryStatus": "DiscoverOK",
        "RedfishVersion": "1.7.0"
    }
}

Discovery troubleshooting

The redfishEndpoint may cycle between DiscoveryStarted and HTTPsGetFailed before the endpoint becomes DiscoverOK. If the BMC is in HTTPSGetFailed for a long period of time, then the following steps may help to determine the cause:
(ncn-mw#) Verify that the xname of the BMC resolves in DNS.
nslookup ${BMC_XNAME}
Expected output:
Server:    10.92.100.225
Address:   10.92.100.225#53

Name:    x3000c0s38b0.hmn
Address: 10.254.1.13
(ncn-mw#) Verify that the BMC is reachable at the expected IP address.
ping "${NODE}-mgmt"
(ncn-mw#) Verify that the BMC Redfish v1/Managers endpoint is reachable.
curl -fk -u root:$IPMI_PASSWORD https://${BMC_XNAME}/redfish/v1/Managers

(ncn-mw#) Verify that the NCN exists under HSM State Components.

cray hsm state components describe "${XNAME}" --format toml

Example output:

ID = "x3000c0s11b0n0"
Type = "Node"
State = "Off"
Flag = "OK"
Enabled = true
Role = "Management"
SubRole = "Worker"
NID = 100006
NetType = "Sling"
Arch = "X86"
Class = "River"

Next step

Proceed to Update Firmware or return to the main Add, Remove, Replace, or Move NCNs page.