[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Referential integrity across reboots
At 03:39 PM 3/3/2003 -0500, Murthy.N.N wrote:
>Hello Jochen,
>
>Can you please elaborate, why such a formula does not work for Logical interfaces?
I agree with Juergen. Many logical interfaces are created in an
implementation dependent order. In general, I believe it is too
late to change the semantics of the ifIndex object now. There
are too many proprietary algorithms for assigning ifIndex values
in use to start mandating a standard algorithm now. It's also
illegal to make such a change, according to SMI rules.
There is a better way to solve this problem. Instead of standardizing
ifIndex assignments, which will be very hard because of all the
different hardware configurations and logical interface initialization
schemes, standardize a read-only "persistent ifIndex" MIB object. If
'true', then the agent will make a best-effort attempt to assign
the same ifIndex value to an interface across a reboot. This needs to
be best-effort because the agent can not always recognize an interface
across reboots, due to HW or SW reconfiguration.
>Thanks and rgds, Murthy
Andy
>Jochen Friedrich wrote:
>>Hi Murthy,
>>
>>> Let us consider an equipment having the entities, rack, shelves, slots,
>>> cards, ports. Let us assume, maximum number of SNMP interfaces (ex; ATM
>>> interface, ADSL line interface, ADSL channel interface) supported per
>>> slot is X. Then, we can device a formula to find out ifIndex of an
>>> Interface uniquely as follows;
>>>
>>> ifIndex=<rack #> <shelf #> <slot #> <interface # within the slot. 0 to
>>> X>
>>>
>>> With this formula, one get same ifIndex all the time for a particular
>>> Interface.
>>
>>That might work for physical interfaces but will definitely fail for
>>logical ones (like loopback, tunnel, ELAN, trunk interfaces etc).
>>
>>--jochen
>
>--
>Murthy.N.N
>
>Local: 7-7927 (NC0. Column Q2-12)
>ESN: 6-357-7927
>External: 1-919-997-7927 (Office)
> 1-919-484-1878 (x 7543) (Res)
>