CRAY System Management - Guides and References > Management Network Switch Rename

Management Network Switch Rename

Any system moving from Shasta v1.3 to Shasta v1.4 software needs to adjust the hostnames and IP addresses for all switches to match the new standard. There is now a virtual IP ending in .1 which is used by spine switches. In Shasta v1.3, the first spine switch used the .1 address. In Shasta v1.4, the ordering of the switches has changed with spine switches being grouped first. The hostname for switches has changed from two digits to a dash and then 3 digits. All IPv4 data for the switches and switch naming comes from Cray Site Init (CSI).

From v1.3, this example system had these IP addresses and hostnames on the HMN network. Similar names and IP address numbers for the NMN and CAN networks as well. Also note that some systems may not have an agg or aggregation set of switches. This is ok. Simply follow the directions below, always paying attention to the CSI specified IPv4 addresses and skip the aggregation switch directions.

10.1.0.1        sw-spine01
10.1.0.2        sw-leaf01
10.1.0.3        sw-spine02
10.1.0.4        sw-leaf02
10.1.0.5        sw-agg01
10.1.0.6        sw-agg02
10.1.0.7        sw-cdu01
10.1.0.8        sw-cdu02

The desired settings for the HMN network would be more like these.

10.1.0.2        sw-spine-001
10.1.0.3        sw-spine-002
10.1.0.4        sw-agg-001
10.1.0.5        sw-agg-002
10.1.0.6        sw-leaf-001
10.1.0.7        sw-leaf-002
10.1.0.8        sw-cdu-001
10.1.0.9        sw-cdu-002

This system needs to do the renames in this order: do CDU switches (8 to 9, 7 to 8) and then leaf switches (4 to 7, but also 2 to 6), then aggregation switches (6 to 5, 5 to 4) and spine switches (3 to 3, and 1 to 2). These have IP address changes and name changes since we now have 3 digits instead of 2 for the switch hostname.

Check switch IP addresses, names, and component names in /var/www/ephemeral/prep/${SYSTEM_NAME}/networks when booted from the LiveCD on ncn-m001.

pit# export SYSTEM_NAME=eniac
pit# cd /var/www/ephemeral/prep/${SYSTEM_NAME}/networks
pit# vi NMN.yaml

Excerpt from NMN.yaml

  ip_reservations:
  - ip_address: 10.252.0.2
    name: sw-spine-001
    comment: x3000c0h33s1
    aliases: []
  - ip_address: 10.252.0.3
    name: sw-spine-002
    comment: x3000c0h34s1
    aliases: []
  - ip_address: 10.252.0.4
    name: sw-cdu-001
    comment: d0w1
    aliases: []
  - ip_address: 10.252.0.5
    name: sw-cdu-002
    comment: d0w2
    aliases: []
  - ip_address: 10.252.0.6
    name: sw-leaf-001
    comment: x3000c0w38
    aliases: []
  - ip_address: 10.252.0.7
    name: sw-leaf-002
    comment: x3000c0w36
    aliases: []

pit# vi HMN.yaml

Excerpt from HMN.yaml

  ip_reservations:
  - ip_address: 10.254.0.2
    name: sw-spine-001
    comment: x3000c0h33s1
    aliases: []
  - ip_address: 10.254.0.3
    name: sw-spine-002
    comment: x3000c0h34s1
    aliases: []
  - ip_address: 10.254.0.4
    name: sw-cdu-001
    comment: d0w1
    aliases: []
  - ip_address: 10.254.0.5
    name: sw-cdu-002
    comment: d0w2
    aliases: []
  - ip_address: 10.254.0.6
    name: sw-leaf-001
    comment: x3000c0w38
    aliases: []
  - ip_address: 10.254.0.7
    name: sw-leaf-002
    comment: x3000c0w36
    aliases: []

pit# vi MTL.yaml


Excerpt from MTL.yaml

ip_reservations:

ip_address: 10.1.0.2 name: sw-spine-001 comment: x3000c0h33s1 aliases: []
ip_address: 10.1.0.3 name: sw-spine-002 comment: x3000c0h34s1 aliases: []
ip_address: 10.1.0.4 name: sw-cdu-001 comment: d0w1 aliases: []
ip_address: 10.1.0.5 name: sw-cdu-002 comment: d0w2 aliases: []
ip_address: 10.1.0.6 name: sw-leaf-001 comment: x3000c0w38 aliases: []
ip_address: 10.1.0.7 name: sw-leaf-002 comment: x3000c0w36 aliases: []

pit# vi CAN.yaml


Excerpt from CAN.yaml showing the two spine switches. Most v1.3 systems would have had these as ending in .1 and in .3. Note these switches are not named "spine" or "agg" since the SHCD may specify differing exit points, but with either option the IPv4 address is specified.

ip_reservations:

ip_address: 10.103.8.2 name: can-switch-1 comment: "" aliases: []
ip_address: 10.103.8.3 name: can-switch-2 comment: "" aliases: []

Save the running-config from all switches before you start.

Note: Mellanox switches require the “enable” command before doing “show running-config”

pit# ssh admin@10.1.0.1 switch# show running-config switch# exit

Save this information in a text file for later evaluation and comparison after all changes have been made. pit# vi before.sw-spine01.txt

Repeat this for all of the switches. The example system has switches up to 10.1.0.8.
Start moves with the highest numbered switch. In this case, that is sw-cdu02. Said another way, if a switch is in a pair, start with the second half of the pair (i.e.. 2 of 2).

Move sw-cdu02 to sw-cdu-002 and increase IP addresses as specified in CSI output. It is a Dell switch.

Note: You can change many addresses in a single session, but not the one you used to connect. This first connection will skip vlan 1 and change all of the other vlans (vlan 2 and vlan 4 on a CDU switch.)
```
pit# ssh admin@10.1.0.6
sw-cdu02# configure terminal
sw-cdu02(config)# hostname sw-cdu-002
sw-cdu-002(config)# interface vlan2
sw-cdu-002(conf-if-vl-2)# ip address 10.252.0.7/17
sw-cdu-002(conf-if-vl-2)# interface vlan4
sw-cdu-002(conf-if-vl-4)# ip address 10.254.0.7/17
sw-cdu-002(conf-if-vl-4)# router ospf 1
sw-cdu-002(config-router-ospf-1)# router-id 10.252.0.7
sw-cdu-002(config-router-ospf-1)# exit
sw-cdu-002(config)# exit
sw-cdu-002# write memory
```