Failed to allocate new Neighbor Table entry

[Jacques Nepveu]

Our 5412zl with firmware K.15.14.0012 sometimes generates the following message:

IpAddrMgr: Failed to allocate new Neighbor Table entry, will LRU oldest host FIB entry (vrf:0 10.4.174.182/32)

What is the neighbor table and how can I query it?

What can cause this message to appear?

Please note that we use static routes exclusively. I am also somewhat perplexed by the IP address in the message since it is not valid for our network.

Thank you in advance

[GM_8]

Have similar problem on a 3500-yl - older software again. (K.14.65) ARP table is filling up with entries from devices outwith the defined subnet (subnet is a /24, the error as in the original post is /32). Why would this cause a loss of connectivity on several independantly defined subnets on the switch? Only "fix" is to clear the ARP cache and connectivity returns. VRF is defined on our core not on edge switches such as the 3500. Very confusing.

[Michael Patmon]

Hello.  IP address manager also manages "local" route (ARP & IPv6 neighbor) entries on the system.  You should only get that log message when you're trying to add an ARP entry and the table is full. When it's full it will try and removed the LRU (Least Recently Used) route in order to make room.  That table is a shared resource with connected/VLAN routes, static routes, and protocol routes. 

So either the table is really full or something else is wrong and the message is printed in error.  How many ARP entries are on the system?  There's no "show arp count" unfortunately so you might have to cut & paste into a text editor.  Also, how many IP routes are in use (show ip route summary)?

As for the second post you should not have to clear the ARP table to get routing working.  Perhaps the LRU mechanism isn't working or routes are not aging out as they should. 

[Jacques Nepveu]

Hello Michael,

The message has not reappeared since I reconfigured a printer usage monitoring software used by a collegue which scanned a couple of 16-bit subnets, generating over a hundred thousand broadcast/arp packets once a day, every day. Of course the great majority of those scanned adresses were non-existant. It seems that the ARP table can be overwhelmed with ARPs for non-existant IPs?!

To answer your original question:

I've been monitoring the only two tables I'm aware of. The forwarding table which holds MAC/Port averages 480 records and the ARP table which holds IP/MAC/Port averages 186 records. All IP entries for both tables are valid IPv4 addresses for our network, we do not use IPv6 yet. I have no idea how many records the ARP table can hold before it gets full.

The result of a show ip route command shows a total of 26 routes, 9 are of type connected and the rest are static.