I see where this has been discussed, but I don't think I saw the
answer to the question which was:
The stats seems to mysteriously reset to 0 on occasion,
No reboots, etc. throws the count off and makes one wonder
how reliable the 'total bytes' stats are.
The counters that track octets are 32-bit unsigned integers. A busy
interface can overflow the largest possible such integer serveral times
a day. When taking successive differences of times series IfOctet data,
you need to test a negative result and add the appropriate constant. In
awk, this looks like:
value = $NF - last
if (value < 0 ) value = value + 4294967296
last = $NF
assuming that $NF is the correct parameter on the line that holds the
counter value.
For every interface, you need to at minimum set the polling rate of the
time series data fast enough that it is physically impossble to roll the
counter twice between polls, i.e. for a full duplex FDDI, if you poll less
frequently than every 340 seconds, you can end up with an unresolvable
ambiguity in your data.
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:14 EDT