Performance issue after upgrade Cluster Switch RCF
Applies to
- BES 53248
- RCF 1.10
Issue
- After Upgrading Cluster Switch RCF to 1.10, Cluster suffered a performance issue
- Cluster Node ports used in Cluster Switch are 0/49 and 0/50
- From RCF file we notice that ports 49-54 needs more manual Steps
config
set clibanner "BES-53248 Reference Configuration File v1.10 for Cluster/HA/RDMA
Switch : BES-53248
Filename : BES-53248-RCF-v1.10-Cluster-HA.txt
Date : 07-20-2023
Version : v1.10
Port Usage:
Ports 01 - 16: 10/25GbE Cluster Node Ports, base config
Ports 17 - 48: 10/25GbE Cluster Node Ports, with licenses
Ports 49 - 54: 40/100GbE Cluster Node Ports, with licenses, added right to left
Ports 55 - 56: 100GbE Cluster ISL Ports, base config
NOTE:
- The 48 SFP28/SFP+ ports are organized into 4-port groups in terms of port speed:
Ports 1-4, 5-8, 9-12, 13-16, 17-20, 21-24, 25-28, 29-32, 33-36, 37-40, 41-44, 45-48
The port speed should be the same (10GbE or 25GbE) across all ports in a 4-port group
- If additional licenses are purchased, follow the 'Additional Node Ports
activated with Licenses' section for instructions
- If SSH is active, it will have to be re-enabled manually after 'erase startup-config'
command has been executed and the switch rebooted"
- Check Additional Node Ports activated with Licenses' section for instructions, we found
!=====Additional Node Ports activated with Licenses===
!Instructions:
! (1) Apply licenses purchased and reboot the switch
! (2) Run 'show port all | exclude Detach', and note the interfaces between 0/16 and 0/55.
! These are the interfaces activated with licenses
! (3) Uncomment the applicable node port license block below, including the 2 exit commands
!
interface 0/49
!no shutdown
!description "40/100GbE Node Port"
!service-policy in CLUSTER
!spanning-tree edgeport
!mtu 9216
!switchport mode trunk
!switchport trunk allowed vlan 1,17-18
!datacenter-bridging
!priority-flow-control mode on
!priority-flow-control priority 2 no-drop
!priority-flow-control priority 5 no-drop
!exit
- After RCF has been installed , found that Cluster Node ports 0/49 and 0/50 has no rules on them
interface 0/49
no shutdown
mtu 9000
exit
interface 0/50
no shutdown
mtu 9000
exit