[rrd-users] AW: Re: problem with manual reset - suggestion

[Pressler Florian, DI]

Von: Alex van den Bogaerdt [mailto:alex at ergens.op.het.net] 

> Update at time "t-1" with "unknown".  Update at time "t" with "0" (you
> *know* the counter is reset, right?).  The next (regular) update will
> have a value and a time so Delta(value) is a known number and
Delta(time)
> is a known number --> no problem.

Aah, that cleared things up a little bit - thank you. I even have a
small improvement to your suggested way of doing it (please correct if
I'm talking nonsense):

I'll clear counters by calling a function which does:

1. Get actual time (z)
2. Retrieve & write actual value of counter to rrd at z-3
3. Reset counter on device
4. Write "unknown" value to rrd at z-2
5. Write '0' value to rrd at z-1

This way should come closest to reality, as there even the transferred
bytes between last regular polling and resetting are taken into account.
The reason why I'd write '0' at z-1 and not z is, because both regular
pollings (1 Min) and resets (1 Month) are run by crontab. If I didn't do
it (i.e., write '0' at z-1), it could be that they are both executed in
the same second, write different values with the same timestamps (z),
which then would perhaps lead to some undefined state.

As far as I have understood the manual of RRDtool, the setting of
heartbeat (in my case 120) should not prevent above theorie from
working. What I did not quite understand is the meaning of xxf. It is
set to 0.5 at the moment - could this lead to a problem?

regards
Florian Pressler

--
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