RRD/Cacti truncating on high bandwidth...

[rkramer]

Just curious if anyone else has had this problem. When some of our links exceed some value above 100mbs, the graphs start truncating the middle portion of the data. They still show the peaks, but drop them down to the bottom...

I've verified the equipment isn't at fault, it is returning the 32bit counter correctly and rolling over every 40 minutes or so.

Anyway, just hoping someone else has an idea... I'm running rrdtool 1.2.11 with cacti 6f.

[rony]

In you data source or data source template, what is the "Data Source Type" set to for the DS's?

[rkramer]

Counter, I had originally been using an unmodified "interface-traffic" template when the problem was first noticed.

here is a debug output from my device rra

/usr/local/bin/rrdtool create \
/var/www/html/cacti/rra/hecn_traffic_in_783.rrd \
--step 300 \
DS:traffic_in:COUNTER:600:0:4294967296 \
DS:traffic_out:COUNTER:600:0:4294967296 \
RRA:AVERAGE:0.5:1:43800 \
RRA:AVERAGE:0.5:24:4380 \
RRA:AVERAGE:0.5:6:17520 \
RRA:AVERAGE:0.5:288:797 \
RRA:MAX:0.5:1:43800 \
RRA:MAX:0.5:24:4380 \
RRA:MAX:0.5:6:17520 \
RRA:MAX:0.5:288:797 \
RRA:LAST:0.5:1:43800 \
RRA:LAST:0.5:24:4380 \
RRA:LAST:0.5:6:17520 \
RRA:LAST:0.5:288:797 \

[bulek]

You cannot use 32-bit counter for measuring such high speeds (if you poll every 5 min). According to your info the counter is rolling over every 40min... how can it be if at the speed of 114 Mbps 32-bit counter rolls over within 5 min?

Anyway you need to use 64-bit counters for speeds greater than 114 Mbps. And one hint for maximum value in your DS settings. You used just maximum 32-bit counter value there. Proper setting should be based on maximum traffic rate in the interface in bytes per sec.

- Piotr

[rkramer]

my mistake, should have done the math! a 32 bit counter will overflow within 5 minutes at 114.53 megabit/sec (with mega=1000, or 111.8 for 1024k/mb) it was overflowing enough that it appeared to not be going over, and seemed like it was incrementing normally...

oh well, now to figure out a lower interval &/or snmp v2...