Hello,
I have a customer who has a stack of two Catalyst WS-C3850-24XS switches which each provide 24 x 10 Gbps ports. The stack is running software version 03.07.03E.
The customer has three VMware ESXi hosts which use VSAN which effectively shares the local storage of each host with the other hosts. The storage updates via multicast on a dedicated VLAN.
The issue I am seeing is the switch ports that connect the ESXi ports in the VSAN vlan are showing output drops and the performance of the hosts is suffering.
I know that the Catalyst 3850 is not the right switch for this type of deployment but I need to try to get it working as well as I can so I have been looking at qos configuration to try to allocate more buffers to the ports.
The configuration I have come up with is shown below:
!
class-map match-any vsan-multicast
description Match VSAN multicast traffic on access list vsan-qos-acl
match access-group name vsan-qos-acl
!
policy-map vsan-traffic
class vsan-multicast
priority percent 100
queue-buffers ratio 100
!
ip access-list extended vsan-qos-acl
permit udp 172.16.100.0 0.0.0.255 host 224.1.2.3
permit udp 172.16.100.0 0.0.0.255 host 224.2.3.4
!
When I try to apply the configuration to a switch port using the commands below they are accepted but do not appear in the show run interface output.
CSWTRE-BF01(config)#int tenGigabitEthernet 1/0/12
CSWTRE-BF01(config-if)#service-policy output vsan-traffic
I tried reducing the priority percentage and queue-buffers ratio to see if that was causing the problem but without success.
Please can you suggest why the service policy is not being applied. I would also welcome comments on the qos config and reducing output packet drops.
Thanks
James
