[rrd-users] Re: graph shooting up

[Mike Wright]

At 17:30 06/08/00 +0800, Mark Manuel C. Ramos wrote:
>    What is the best way to solve the problem of graphs shooting up when the
>device being polled goes down for a while (let say 30 mins).  What value
>should I use to update the RRD dbase if cannot reach my device.

If you can't contact the device you are polling, then you should log
Unknown. You should also read about the "x-files-factor" to see how this
works, although using an xff of 0.5 should be OK. 

It is important that you set the minimum and maximum settings for your
interfaces or you will get spikes. Once you have tuned your rrds, you can
remove the spikes using the rrdtool dump / rrdtool restore --range-check
functions.

At a more advanced level, you can detect and compensate for reboots in your
data collector too. When you poll the device, check the sysUpTime. If
$sysUpTime < (time - $lastupdate($rrd)) then there has been a reboot since
your last update.

Assuming you are logging data for a COUNTER type DS, you should log an
Unknown value at the sample time "time - sysUptime" (ie the time of the
reboot), then log a 0 for the sample time - sysUpTime + 1 second. Then log
the actual value of the counter at the present time. This will prevent
spikes in your graphs and also allow you to log sensible data even when a
reboot occurs between updates.

Good luck!

Mike Wright



--
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:rrd-users-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/rrd-users
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi