cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
2780
Views
0
Helpful
7
Replies

Broadcom LiveLink : Receiving MAC flaps with Cisco Nexus 7000

djgeorges
Level 1
Level 1

We are migrating from using two Nortel 8600's running VRRP at the distribution to Cisco Nexus 7K's using HSRP.  So we have a server connected to two 3750G switches which then connect to the Nexi (previously the 8600's).  As soon as we connected the 3750's to the Nexus and moved the gateway to Nexus, LiveLink forces all the servers to alternate traffic between NIC1 and NIC2. 

Since LiveLink is a teaming application, it uses virtual mac for nic1 and nic2, but the virtual mac associated with the IP address moves to the active link.

LiveLink is used to check the availability of the gateway by polling the gateway out of each interface using an ARP request.

The problem does not exhibit itself in our Cisco VSS environment, and with Nortel's VRRP.  I tried running VRRP on the Nexus but no joy.

Anyone know of a bug that could cause this issue?

7 Replies 7

Jerry Ye
Cisco Employee
Cisco Employee

LiveLink is causing unicast ARP flooding to the gateway.

The resolution I've seen so far is not to use the LiveLink.

HTH,

jerry

Unfortunately we have LiveLink enabled on most of our Windows servers in our data centers.  One of my colleagues sent me this bug issue.  I'm not sure if this is the cause, but it's worth trying.   We will update the NxOs (currently on 5.1.1) next week and see if that fixes the problem.

•CSCtl85080

Symptom: Incomplete Address Resolution Protocol (ARP) entries are observed on a Cisco Nexus 7000 Series switch, along with partial packet loss and a memory leak.

Conditions: This symptom might be seen when ARP packets have a nonstandard size (that is, greater than 64 bytes).

Workaround: This issue is resolved in 5.1.3.

I don't think this is the bug. You can give it a try but please be aware that 5.1.3 has a kernel memory leak issue. The bug ID is CSCtq62339.

Regards,

jerry

The problem has been resolved. 

We upgraded from 5.1.1 to 5.1.4.  I'm going to hear about this from the Intel and AIX guys for weeks

Wait, so is this fixed by 5.1.4? Or the server guys changed the config on the Broadcom driver?

Upgrading the NxOS fixed the problem. 

Our server guys wouldn't touch this problem with a ten foot pole.  Their take on it was "If it worked before the Nexus was deployed, you should get rid of the Nexus."

Thanks, good to know though.

Review Cisco Networking for a $25 gift card