From: "Stephen J. Roznowski" <sjr@home.net>
On 17 Jul, Matthew Stier wrote:
> I don't think the "spikes" are Tobin's issue. I suspect that Cricket is the
> cause here, compounded by the fact that what Network Appliance has in there
> MIB is not typical units. Most systems report "octets". The NetApp is
> reporting kilobytes of traffic. Since Cricket is assuming octet counts,
> when the counter rolls over, Cricket is trying to compensate by
> "guessing" the counters maximun value, and correct the received value for
> the roll over.
>
> What we need from Jeff, is a method to either set a maximum value for a
> counter, on a target by target base, or a means to scale the incoming values
> on SNMP obtained data, without having to resort to external scripts.
My brief reading of the code, leads me to believe that the counter
rollover is occuring in rrd_update.c (part of rrdtool). I guess that
this could be fixed in cricket, by allowing the value to be scaled (in
this case by 1024) before storage....
-SR
--------------------------- ONElist Sponsor ----------------------------
Congratulations MilitaryEFM. Our latest ONElist of the week.
http://www.onelist.com
How is ONElist changing YOUR life? Visit our homepage and let us know!
------------------------------------------------------------------------
This archive was generated by hypermail 2b29 : Mon Mar 06 2000 - 19:00:56 PST