[j-nsp] SNMP polling a logical-system without encoding it asLS/RI at community

[Zenon Mousmoulas]

On 15 Ιαν 2009, at 1:44 ΜΜ, Subodh Kumar wrote:

> It would depend on which interface is your packet entering the system
> (router).  Consider an example:
>
>            ________                    _________
>           |       |                    |       |
>           |       |      (lr1/ri2)     |       |
>           |   R1  |--------------------| nms1  |
>           |       |fe-0/0/3        em1 |       |
>           |_______|                    |_______|
>
>
>   - (lr1/ri2) indicate the routing-instance
>   - Ip Address:
>        nms1 (em1): 14.14.14.1
>
>
> logical-systems lr1 {
>   routing-instances {
>      ri2 {
>             ...
>             interface fe-0/0/3;
>             ...
>      }
>   }
>   interfaces{
>      fe-0/0/3 {
>           unit 0 {
>             family inet {
>                address 14.14.14.2/24;
>             }
>         }
>      }
>      lo0 {
>        unit 1 {
>           family inet {
>               address 14.14.14.100/32
>           }
>        }
>      }
>   }
> }
>
>
> Now if you send a SNMP packet from 'nms1' on ip-addess 14.14.14.100 or
> 14.14.14.2, that makes sure that packet is entering the system from
> inetrafce fe-0/0/3 which is part of logical-systems 'lr1'. In this  
> case
> you don't need to specify the prefix 'lr1/default' in the community
> string.
>
> In your case, it looks that the SNMP packet is getting into the router
> through 'fxp0' or some interface that is part of 'default'
> routing-instance. Does the 'ping 10.0.0.5' from the NMS work for you?


Yes I can ping from the nms to the loopback of the logical system, but  
you are right, traffic is routed between nms and lr1 through R1 = it  
enters through an interface that belongs to R1 (the physical router).

Thanks for the suggestion! This looks promising. We will slightly  
alter the network topology, in order to try this out.

One question though: I see that you mention ri2, which is different  
from the default routing instance. Does that actually matter? So far  
we have not been using routing instances, but do you think we actually  
need to setup one explicitly, within the logical-system?

Best regards,
Z.