Ether Channel from 6513 to F5 Load Balancer

[jim.totman]

I have an F5 Load Balancer connect to a 3750 via an ether channel that contains two gig ports. It works great, but when I copy the port configurations to my 6513 I get input errors.

The 3750 is running Version 12.1(14)EA1a and the 6513 is running 12.2(17d)SXB1.

Any help would be appreciated.

Here is the configuration.

interface Port-channel1

no ip address

speed nonegotiate

switchport

switchport access vlan 137

switchport mode access

switchport nonegotiate

spanning-tree portfast

end

interface GigabitEthernet5/23

no ip address

load-interval 30

speed nonegotiate

switchport

switchport access vlan 137

switchport mode access

switchport nonegotiate

no cdp enable

spanning-tree portfast

channel-group 1 mode on

end

[pflunkert]

Are you sure that you don't have connectivity problems? When you see input errors it could be an error with the cable or the GBic. Had you tried another cable and/or port?

You configure a manually etherchannel. Had you tried a LACP Etherchannel on the catalyst? (channel-group 1 mode active)

Regards

Peter

[jim.totman]

The channels do not come up with the active mode statement.

F5's Big IP do not support LACP.

I am still getting the errors on the 6513, but I do not get any on the 3750 switch.

These devices are directly connected to the 6513 via fiber. The connections run over multiple blades and all the ports are getting the errors.

I do not think it is a physical connection issue.

[pflunkert]

Hi,

which module you use on the cat 6509? Had you tried another module?

Regards

Peter

[jim.totman]

The modules are WS-X6724-SFP. All bundles have half its ports on one card and the other half on the other card.

We disabled the port channel yesterday and used a single port for connectivity. We did not get any errors. Once we put the port back into the port channel and reenabled it the errors came back.

We have captured the 6513 putting arp request that are only 60 bytes. Does the 6513 record these as errors?

Jim

[pflunkert]

Hi Jim,

sorry for my late response. I would made a test in our lab. But unfortunally i had many things to do.

I found a bug which describe the similar behaviour you see:

CSCec07464

On the catalyst 4000 family IOS switches, input errors may incrment for an interface. This does not always indicate any loss of data traffic.

Sometimes, symbol errors and sequence errors increment for an interface but do not result in corruption of data packets. Hence the problem seems to be cosmetic.There is no workaround for this problem.

The problem has been resolved in 12.1(20)EW, 12.1(20)E and later releases. In these releases, symbol errors and sequence errors that do not affect data packets are not counted as input errors. Symbol errors for an interface are counted and can be seen

with "show interface counters detail" command. They may indicate a problem with the physical link.

This problem is related to your software also (12.2(17d)SXB1). Perhaps you should try anther software.

When this don't solved your problems, i would open a tac case.

Regards

Peter