[mrtg] Re: Average's messed up..

[Alex van den Bogaerdt]

Jeff  Warnica wrote:
> 
> Yep, the latest and greatest..
> http://landon.chebucto.ns.ca/mrtg/hfxlogins.html graphs tty (modem) and pty
> (telnet) users. The monthly gives a Max Telnet of 23 and the yearly a max of
> 11. The monthly graph isnt yet full. Im just not sure if I can trust the
> summaries untill the graphs are full.
> 
You may want to play a bit with the config file.  You can alter the various
labels.
"Bytes per second" -> "connections"  using Ylegend
"B/s" -> "" using ShortLegend.


The "max" number is the average from the interval with the highest average.
It isn't the absolute maximum seen.

T+5min   20 users
T+10min  10 users
T+15min  10 users
T+20min   5 users
T+25min   5 users
T+30min  10 users
Daily:  Average: 10 users, Max: 20 users
When the samples are compressed to fit in the weekly log:
(20+10+10+5+5+10)/6 = 10 users max and avg.

Try "WithPeak[^]: ymw" at the top of your config file.
Doing so you may also want to play with the colors (Colour[])



The averages are also okay:
day:   9 and 22
week:  6 and 20
month: 7 and 21
year:  7 and 20

Close enough ...
During the evening you have more users than on average.  The daily graph
shows two evenings and only one morning.

> These are generated with the gauge option, I see that on my pm ether pages
> that all seems sane.
> 
It does.

> 
> I dont know how hard it would be to implement, but a option to ignore points
> with no data when caculating the text summaries might be usefull.. Users may
> have a auto dialing ppp router (just there workstation, or a lan with a dial
> up link).. If the average bandwidth while the link is up is over some
> threshold then there might be a case for a faster link. If the users aren't
> maxing out the link when it is active, then they might not have a case for
> upgrades.
(they? or you? :-)

What you want to monitor is how many seconds the line is above a certain
treshold.  Don't use averages for that, write a script that does this job.
You'll have to use the external program support for this.

Just as something to start:
Update your ethernet stats as usual.  The second line in the mrtg.log file
will hold the current rate.  If the rate is above a certain level, return
the number of seconds lapsed (difference of line 2 and 3, column 1).
If the rate is below that limit, return 0.
One of the thresholds is 0 (to signal activity) and one could be 90% of
the link speed.

This may need some work but I hope you get the picture.


regards,
-- 
   __________________________________________________________________
 / alex at slot.hollandcasino.nl                  alex at ergens.op.het.net \
| work                                                         private |
| My employer is capable of speaking therefore I speak only for myself |
+----------------------------------------------------------------------+

--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg