[e-nsp] old, summit400 (ExtremeWare) and L3 routing concept

[Marcin Kuczera]

hello,

today I had to do some troubleshooting and I couldn't understand L3 IP 
Routing in ExtremeWare (summit400).

Documentation specifies that it has:
Forwarding Tables
• Layer 2/MAC Addresses: 16K
• Layer 3 forwarding database in hardware: 4K
• Layer 3 routing table size: 64K


so, ipfdb - 4k

but, it in newer EXOS - there is nothing like ipfdb, just a routing table.

ipfdb looks like a table with bindings - DST IP - NextHop MAC.
And it is created for every signle DST IP that the switch is forwarding 
traffic.

So, I have observed, that if there is a lot of traffic towards different 
destinations, it is very ease to make ipfdb full and to get:
06/14/2014 08:40:07.20 <Erro:IPHS> Could not add L3 entry for unit 
3-93.105.223.170 at 0xbff(-6: Table full)
06/14/2014 08:40:07.20 <Erro:IPHS> Could not add L3 entry for unit 
2-93.105.223.170 at 0xbff(-6: Table full)
06/14/2014 08:40:07.20 <Erro:IPHS> Could not add L3 entry for unit 
1-93.105.223.170 at 0xbff(-6: Table full)
06/14/2014 08:40:07.20 <Erro:IPHS> Could not add L3 entry for unit 
0-93.105.223.170 at 0xbff(-6: Table full)

What happens than ? CPU forwards the packet ?
What is corelation between routing table size (64k is not that small) 
and forwarding table ?

It there any way to install somehow some entries in ipfdb that will 
cover particular networks ?

is it that command:
configure ipfdb route-add clear-subnet

default is clear all.

Sorry for asking silly-like questions, but this is wired for me...

Regards,
Marcin