The description I heard is that as long as there are no interfaces
faster than 20Mb/s that 64-bit counters won't be used.
Don't know how accurate this is as i'm no uberalles snmp guru.
- jared
On Tue, Feb 12, 2002 at 06:50:13PM +0100, Gert Doering wrote:
> Hi,
>
> (copying cisco-nsp back in, maybe we'll manage to warn a few other readers
> of the pitfalls... or annoy Cisco enough so that they go and fix it)
>
> On Tue, Feb 12, 2002 at 08:10:52PM +1300, Andrew Cutler wrote:
> > > <rant>
> > > How broken can this get???
> > Lots more. I have some 4500 that I recently started graphing with V2 and
> > couldn't figure out why the graphs looked the same. So I started
> > experimenting with walks and gets. The active interfaces are eth0(1),
> > serial0(3) & serial1(4) Check out the following:
>
> Oh, yes, we discovered that one quite a while ago.
>
> Some platforms (if I remember correctly, it hits only the 4500) do not
> implement counter64s, but don't tell you so - if you query the counter64,
> it will always return the last counter32 that you have queried before.
>
> Or maybe it was "if you query the counter64 on an *interface* that doesn't
> support it, it will return the last counter64 value from the last c64-capable
> interface queried".
>
> So our tool has an check for IOS versions, and falls back to counter32
> on those platforms (which is no big problem for us as those are low-speed
> customer aggregation boxes).
>
> [..]
> > V1 snmpwalk the interfaces (works fine):
>
> Yep...
>
> > This was all on 12.0(19). So we set up another 4500 and tried it on
> > 12.0(20), 21 and 21a with identical results. Then, just out of
> > curiousity we tried 12.2(7) which came back with "no such object in the
> > mib" for ifHCInOctets (does this mean 12.2 doewsn't support V2?). At
> > this point we were so flabbergasted we gave up.
>
> Counter64s are only supported on high-speed interfaces, not on "-T"s,
> so our tool always queries both, and if no Counter64 exists, or if
> it's 0, the Counter32 will be used.
>
> Except on ATM subifs, of course, where it will *always* use Counter64,
> as "0" could also mean "has just been resetted by 'clear count'" (yuck)...
>
> > The good news is that this particular bug only seems to be happening on
> > 4500's. Our 75/7200's don't seem to be affected, but I'll have to check
> > the details of your message on them.
>
> I have not seen this on any other platform or any other interface. Just
> this single PA-E3 on this single 7206VXR/NPE-400, 12.0(19)S1...
>
> 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
-- Jared Mauch | pgp key available via finger from jared@puck.nether.net clue++; | http://puck.nether.net/~jared/ My statements are only mine.
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:33 EDT