Nexus 7000 - Adjacency table exhausted

[n.poongsawad]

Hi Experts,

Please give us some suggestions. I'm running a couple N7K as campus's core. Separate two LC on each box to TWO VDC e.g. Backbone & Wireless VDC. Wireless VPC is paring to 4 WLC, 2 links each. Between N7K and WLC running in layer 2 mode. We serve both IPv4 & IPv6.

Problem is one day on last week, the first day of university's semester. The N7K goes corrupt. After along while troubleshooting we found that due to this logs

N7K1 %IPFIB-SLOT4-4-CLP_FIB_ADJ_EXHAUSTED: Adjacency allocation failed on instance 6

We have tested for how adjacency table being accumulated. Then we found the following captured picture.

What I did wrong? It seem like N7K F2e Line-card be able to hold only 10,000+ MAC address (host)?

Please give us some suggestions.

Nipat CCIE#29422

[plustgraaf]

The F2e cards can only support 16K MAC addresses per SoC (Switch on a Chip) and a total of 16K ARP entries total.  The command "show hardware capacity forwarding" will show you the complete picture for layer 2.   That's why we now have 7706s with F3 cards.

[sbhadrav@cisco.com]

Some configuration mismatches between vpc peers can cause unexpected or wrong forwarding.Therefore some checks are
performed on the configuration of the vpc  peers to see if those parameters are matching, these parameters can be global or
on a vpc. The configuration parameters that need to match on both vpc peers are classified as type-1 and type-2.Type-1
mismatches being more severe,trigger some actions from the devices, such as suspending a vpc, not allowing the cluster to form.
With type-2 configuration error, an error will be syslogged, but no action will be taken by the nexus.For complete list of parameters
that must or should be identical between the vPC peers refer Compatibility Parameters for vPC Interfaces

Note that  Note that Interface vlans must be in the same state (admin up or down) on both vPC peers.

To display the consistency of parameters that must be compatible across the virtual port-channel (vPC) interfaces,
use the show vpc consistency-parameters command. .

[mmilneatcarleton]

I'm going through the same thing, pretty much identical to you, wireless blew up on the first day of school.

The earlier assessments are correct, we are hitting the 16K limit on the F2 line cards. In the short term, you can drastically cut down the table usage by disabling IPv6 on the wireless VLANs. This will really help as each IPv6 client has two adjacencies that must be tracked, one for the unique global IPv6 address, and another for the link-local address. I've disabled IPv6 but I'm still bumping against the limit during peak usage times.

The long term solution is F3 or M3 line cards, but even then you must be careful mixing F2 and F3 line cards across VDCs.