ARP entry constantly listed as imcomplete

UHansen1976 · ‎12-22-2010

Hi,

I'm faced with bit of a problem here.

I have a pair of Cat-6509/Sup720 running a dist.layer switches. The two switches are Layer2 adjacent across an etherchannel and running rpvst. Each dist.switch is configured as root bridge for half the vlans and secondary for the other half, and vice versa.

Some 100+ vlans are configured, using hsrp as fhrp. A pair of N5K-switches are attached to the Cat6509's, acting as 10Gig accesslayer switches. Among others, an IBM Power7 server is attached with 4x10G trunk.

One vlan in particular is causing me a problem. The vlan appears to be working corrently, no misconfigurations have found on either the SVI, Hsrp or spanning-tree. However, a few arp entries are constantly listed as incomplete on the one secondary root bridge. Any icmp-echo packet, that happends to arrive on the secondary root, experiences timeout. However, if it ends on the other switch, it responds instantly.

9/10 arp entries are cached flawlessly, but a few remain in the incomplete state. This however is not the case on the primary root. I've tried to clear the arp-cache and perform a shut/no shut on the svi, but nothing resolves the problem. The same ip.addr are listed as incomplete, when ARP tries to resolve them.

Any help is greatly appreciated.

Thanks and happy holidays

/Ulrich

Reza Sharifi · ‎12-22-2010

Hi Ulrich,

Have a look at this doc. It discusses what may cause incomplete ARP entries:

https://supportforums.cisco.com/docs/DOC-2094

HTH

Reza

UHansen1976 · ‎12-23-2010

Hi Reza,

Thanks for your reply.

I did actually look at the document prior to troubleshooting the problem, but where unable to come any closer to a solution.

The thing is, that all but a few MAC's, are cached correctly and can easyli be accessed from the secondary root-bridge. The hosts are all virtual hosts, hiding behind one physical NIC and currently 7 out of 8 virtual hosts can be accessed from either of the two root bridges. So I've ruled out any vlan- or spanning-tree misconfiguration, since all other hosts are working just fine.

/Ulrich