[c-nsp] ifindex persistence

[Andre Beck]

Hi,

On Tue, Apr 12, 2005 at 12:36:18PM -0400, christopher_a_kane at bankone.com wrote:
> on the network management side of things - has anyone employed the use of 
> snmp-server ifindex persistence?

Of course. It's extremely valueable when you are monitoring your network
with software that cannot dynamically adapt to changing ifIndexes, which
is the majority. Even if it can, this may involve extreme overhead and
thus having a persistent ifIndex mapping is great anyway (i.e. of course
you *could* run a Torrus rediscover after every router reload, but not
having to do so spares you lots of CPU-seconds [if not minutes] and
plenty of SNMP walking traffic [including the CPU load involved by that
on the network gear]).

Newer switch plattforms just use highly offset fixed numbers for their
hardware interfaces what makes persistence way easier.

> success/horror stories?

Mainly success as long as you stay with one IOS version train. It
survived any upgrade within 12.0S so far, for instance. Then again,
going from 12.0S to 12.2S entirely hosed it which is understandable
as the interface naming and representation in the ifTable changed
significantly, but nevertheless, it sucks to have to redo all your
tkined, ifIndex-related Nagios (including the memory pool stuff as
that changes slightly on that upgrade) and legacy MRTG feeds. That's
where Torrus comes in handy - just a rediscover. Sadly, the memory
pool stuff hoses Torrus too, as it names the RRD data sources with
the index number included and one pool changes its index. Can be
fixed with rrdtool renaming the DS to what Torrus expects, though.

HTH,
Andre.
-- 
                  The _S_anta _C_laus _O_peration
  or "how to turn a complete illusion into a neverending money source"

-> Andre Beck    +++ ABP-RIPE +++    IBH Prof. Dr. Horn GmbH, Dresden <-