Calculating for MAX

[mihael]

I'm trying to manually calculate for the MAX value of my interface via python script, but I can't get the same result as the one shown in my cacti graph.

Script:

Code: Select all

data = rrdtool.fetch(rrd_file, 'AVERAGE', '-r', '300', '-s',  '1588069832', '-e', '1588674632')
    for row in data[2]:
        if row[1] is not None:
            max_out.append(float(row[1]))
 max_out.sort()
 max_value = max_out[-1]
 max_bps = max_value * 8
 max_mbps = round((max_bps / 1000 / 1000), 2)
print(f'{max_mbps}M)
#Result: 591.27M

Cacti Graph:





I even tried manually running the rrdtool graph command, but it's also giving me a different result:

Code: Select all

➜  rrd_files rrdtool graph -s 1588069832 -e 1588674632 \
DEF:a='1min_traffic_in_14983.rrd':'1min_traffic_in':AVERAGE \
DEF:b='1min_traffic_in_14983.rrd':'1min_traffic_out':AVERAGE \
CDEF:cdefe='b,8,*' \
PRINT:cdefe:MAX:'Maximum\:%8.2lf %s'
0x0
Maximum:  538.64 M

Is there any cacti calculation or anything that I'm missing causing the graph to display a different result?

[cigamit]

trying changing
max_mbps = round((max_bps / 1000 / 1000), 2)
to
max_mbps = round((max_bps / 1024 / 1024), 2)
and see if it gets you a bit closer.

[mihael]

trying changing
max_mbps = round((max_bps / 1000 / 1000), 2)
to
max_mbps = round((max_bps / 1024 / 1024), 2)
and see if it gets you a bit closer.

Yeah it does, now it's giving me 563.88M.
I've also tried calculating the 95th Percentile, which I was able to accurately do and I was using 1000 to convert bits to mbps/gbps

[TheWitness]

Look at the base value of the graph template.

[mihael]

Look at the base value of the graph template.

But shouldn't the cli command `rrdtool graph` have the same result for that as it is the same command shown by cacti?
Also, I checked the "Base Value" for the graph, it's '1000'.

[TheWitness]

I thing the problem is Python, lol. Just kidding. This is a good question for Tobi. It may simply be that RRDtool chose to pull the data from another less/granular archive that was different to what the fetch function choose. The Python math looks good to me.

[TheWitness]

Looking again, you need to ensure you are consistent with resolution. In your Python call, you are calling out the 5 minute rra, Cacti let's Tobi pick.

[mihael]

Looking again, you need to ensure you are consistent with resolution. In your Python call, you are calling out the 5 minute rra, Cacti let's Tobi pick.

I tried calling out the one minute rra, and it is giving way more max value.

[TheWitness]

Well, once you figure it out, report back here, some people are lurking.