SNMP counter64s gettings stuck?

From: Gert Doering (gert@greenie.muc.de)
Date: Thu Feb 07 2002 - 05:07:04 EST

Next message: pankaj: "sorry test message"
Previous message: Desmarais, Jonathan: "RE: [nsp] VIP4-80/PA issues, anyone?"
Next in thread: Gert Doering: "Re: SNMP counter64s gettings stuck?"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Mail actions: [ respond to this message ] [ mail a new topic ]

Hi,

got a new one today... the fact that "show interface" output counters
sometimes get stuck and stop counting until one does "clear counter"
is well-known and is plagueing us since ever.

Today I had an SNMP counter64 get stuck on me... 7206VXR/NPE-400,
IOS 12.0(19)S1, PA-E3. Those are the values read via SNMP over the
last hour...:

20020207|1032|333000|2356000|20421938485|95908295589|up|up|sp1
20020207|1037|543000|2668000|20421938485|95908295589|up|up|sp1
20020207|1042|311000|2203000|20421938485|95908295589|up|up|sp1
20020207|1047|516000|1708000|20421938485|95908295589|up|up|sp1
20020207|1052|731000|2000000|20421938485|95908295589|up|up|sp1

the 3rd+4th column are "5 minute average" (so there is really traffic
flowing), the 5th+6th column are the ifinbytes64/ifoutbytes64 values
($MIBii.1.31.1.1.1.6 and .1.31.1.1.1.10).

Interesting enough, the 32 bit counters continue to increase as expected.

So, now my question: has anybody seen this before? Is this a known bug?
Is it fixed in 12.0(21)S? Is there any chance to get these counters from
hell fixed one day?

<rant>
How broken can this get???

How can the lower 32bit half of a counter continue increase and the 64bit
counter get stuck? How many independent-and-buggy counters have they
implemented there?

We're currently using "show int" AND "snmp" values for "5 minute average"
and for "in/out octetes", and compare all those values to each other
regularily to see what funny new bugs have crept up - and this is really
ridiculous (for example the "5 minute average" from "show int" and from
"SNMP" vary up to 20% if averaged over a full day - which should never
happen as it should be the *same* counter).
</rant>

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@greenie.muc.de
fax: +49-89-35655025                        gert.doering@physik.tu-muenchen.de

Next message: pankaj: "sorry test message"
Previous message: Desmarais, Jonathan: "RE: [nsp] VIP4-80/PA issues, anyone?"
Next in thread: Gert Doering: "Re: SNMP counter64s gettings stuck?"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Mail actions: [ respond to this message ] [ mail a new topic ]

This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:31 EDT