[c-nsp] Monitoring L3 link status

[Frank Bulk]

Lots of good ideas coming, but I like the SLA one the best.  It's cleanest,
routing wise, because I don't need to do any SBR or any route modifications.
It also doesn't depend on the state of BGP, which, while important, doesn't
tell me the availability of the far side.  And the OID of
'rttMonReactOccurred' seems to be what I need to poll.

I like BFD, too, and it seems to be supported by Cisco and Juniper, but I'm
not sure if it's on the 3750ME (a quick google mentions that it was
introduced in release 12.4(9)T, and we're running 12.2(25)EY4).  I also
don't see a SNMP MIB ready and available.

Regards,

Frank

-----Original Message-----
From: Phil Mayers [mailto:p.mayers at imperial.ac.uk] 
Sent: Wednesday, December 13, 2006 7:30 PM
To: frnkblk at iname.com
Cc: Cisco-NSP Mailing List
Subject: Re: [c-nsp] Monitoring L3 link status

Frank Bulk wrote:
> We're using two Cisco 3750-ME routers (B & C) that are connected to our
> upstream service provider's routers (D & E).  We started using ping checks
> to D & E from our server running SolarWinds, but because our internet
> traffic runs primarily over B-D or C-E it's not really a good link test.
If
> our traffic runs over C-E then ping checks to D run over C, E, and then on
> to D.  Link B-D could fail and we wouldn't know.
> 
>            |---OSPF---[B]---BGP---[W]---[EoS]---[EoS]---[D]
>            |           |                     / \         |
> server ----A         OSPF                   |   |       BGP
>            |           |                     \ /         |
>            |---OSPF---[C]---BGP---[X]---[EoS]---[EoS]---[E]
> 
> This can't be a new issue -- what's the easiest way to test L3
connectivity
> between B-D and C-E?  All our connections internally and to our upstream
> provider are Ethernet, transport is Ethernet over SONET) and testing for
> interface status is not reliable enough as our demarc with our upstream
> provider is a Cisco 2950 (W & X).  I believe our upstream provider uses
> Juniper routers.

For probing / monitoring, the ip sla / rtr functionality from B-D and 
C-E would be the normal way (assuming D and E answer pings). It doesn't 
run continuously and getting the results back via SNMP can be a pain, 
but it's a reasonable solution.

http://www.cisco.com/go/saa

For actually detecting failures on the link, something like BFD or 
802.3ah would be the correct solution. I don't know if the 3750ME 
supports either, and the latter is quite new and still therefore raw.