[c-nsp] IPv6 ND cache via SNMP

[Phil Mayers]

On 10/19/2010 01:03 AM, Michael Sinatra wrote:
> Is anyone out there polling the IPv6 neighbor discovery cache via SNMP?

Previously, yes. I get them via expect/cli now, because the OID sorting 
required for snmpwalk of that table on 6500s is prohibitively expensive 
when it gets very large (well - it is for IPv4 & ipNetToMedia; I am 
assuming the same for ipv6, and since the expect script already runs for 
v4...)

>    I am mainly interested in getting the cache from 6500s running SXI4a
> on the VS-720-10GE-3C.  In earlier IOS versions (on different platforms,
> I believe), this was done using the interim CISCO-IETF-IP-MIB
> (specifically cInetNetToMediaTable), but it seems as though this should
> all have been merged into the new RFC 4293-compliant IP-MIB.  However,
> with ip.ipNetToPhysicalTable, I get 'no such object'.

Not yet I think.

> ipv6NetToMediaTable (part of the IPV6-MIB) works great on JunOS, but not
> on cisco (also 'no such object').  It's not clear from the MIB locater
> if this is even supported in SXI4a--looks like not.  Are we really still
> that far from IPv4/IPv6 feature parity?

Shrug. I wouldn't read too much into it.

Our code tries:

CISCO-IETF-IP-MIB::cInetNetToMediaEntry
1.3.6.1.4.1.9.10.86.1.1.3.1

...this still seems to work on our SXI4a test box (I just tested it); 
remember you won't see anything if the neighbour cache is empty, as it 
often is on quiet test boxes (I find).

> Currently what I am doing is scraping "show ipv6 neighbor" via RANCID
> and shoving it into a flat file for processing and insertion into a SQL
> DB.  But...yuck!  This would be a lot cleaner with SNMP--and far fewer
> moving parts.  One perl script could easily poll and push into SQL all
> at once.

Well, as I say above - that approach has advantages.