traffic issue on Cat6500 switch

[HWangLoyalty_2]

We have two Cat6500 switches (core1 and core2)interconnected with trunk port. we also have a few of access switches to uplink to two core switches with trunk port.In core switch, we use SPT (RPVST)for L2 load balancing and HSRP for L3 load balancing according to odd and even Vlan. For example,root bridge of SPT and active VIP of HSRP of odd vlan are in the core1 Otherwise, it is in the Core2. My question is the traffic would go through two core switches when even vlan would try to access odd vlan,right?

There is a backup server in the odd vlan. we need backup all of data to that server during the midnight. I suspected I should see the huge traffic in two core switches when other hosts in the even vlan tried to access odd vlan. But so far the huge traffics I only saw were in core2 switch.

Please advice. I would apprecaite it!

[Giuseppe Larosa]

Hello Huan,

my first thoughts are the following:

I would check if ip proxy-arp is disabled on the two core switches SVIs.

servers relying on proxy-arp could get the physical MAC address of core2 to reach the backup server.

Another aspect can be icmp redirect:

is the odd vlan defined also on core1 ?

otherwise if core1 uses core2 to reach the backup server it could send out icmp redirects and the traffic can be handled by core2 only

Hope to help

Giuseppe

[HWangLoyalty_2]

Thanks Giuseppe.

We did not disable "proxy-arp" on the two core switches SVIs. I also could find arp entry on the core2 switch. So you mean if servers get the mac-address of backup server from core2, it would not query it to core1, right?

If I want to fix it, do next step:

a. disable proxy arp in the backup SVI

b. clear arp belong to backup server vlan.

Please advice!

icmp redirect is also enabled. But I do not think it caused that issue.

Thanks again.

[Giuseppe Larosa]

Hello Huan,

you need to disable proxy-arp on the vlans where the servers that need to talk to backup server are connected

to be noted they should rely on proxy arp only if they miss the default gateway or they have an incorrect less specific network mask that make them to try to ARP for the backup server ip address

icmp redirect can play a role only if the subnet of the backup server is defined only on core2 that is unlike.

Hope to help

Giuseppe

[HWangLoyalty_2]

OK!

I also found the broadcast traffics were increased quickly with "show interface (physical or SVI)" when huge traffics got through our core switches. Does it cause the command "icmp redirect" enabled? Thanks again!