[rrd-users] Re: About logging the REAL valuesn
don.baarda at baesystems.com
Wed Aug 23 02:40:05 MEST 2000
> -----Original Message-----
> From: Alex van den Bogaerdt [SMTP:alex at slot.hollandcasino.nl]
> Sent: Wednesday, August 23, 2000 8:36 AM
> Enrique Sanchez Vela wrote:
> > I dont mean to be a smart a.. but I think Tobi made
> > this very clear, the MIN and MAX values are calculated
> > out of the averaged values, if we dont like it or does
> > not work for us and Tobi does not agree on changing
> > it, then we have two choices here, either stop using
> > rrd or fix it for local use.
> This is the whole point of discussion. It may or may not lead to
> changes in the program. Without this discussion, nothing will ever
> change. There is an easy way to work around the problem, but for most
> new users this is not clearly visible.
This is probably hits the nail on the head. At my first glance, I
was concerned that averaging would interfere with my raw data. After
thinking about it more, I realised averaging made sense, and I probably
didn't want the raw data.
> Remember, this is a discussion list, not a helpdesk. I started the
> topic off-list so obviously Tobi agreed that is was worth a discussion...
Perhaps a "RAW" DS type would be useful for some applications. This
would simply store the raw sample value for every entry within heartbeat
time since the last sample. It would be like a "gauge" DS, except assumes
that the gauge stays constant between samples. Note that this would mean
only the last sample would be used if multiple samples were recorded in a
single step, but I think this is what happens with the other DS's anyway.
In theory it would also be possible to give meaning to MIN, MAX, and
LAST CF's for RRA's with steps=1 when there are multiple samples taken
during a single step. However, this would require saving additional data for
each DS, and if you really want this type of behaviour you should probably
just reduce your step size. It would also be totally meaningless for
anything other than GAUGE (or RAW) DS's (the other DS's record counts/sec,
which means you need a common time interval to compare them meaningfully).
The only other possibly useful thing would be a 'FIRST' CF to
complement 'LAST'. Don't share price graphs often show opening and closing
Personally, it does what I need, and has been pretty well thought
out. If it doesn't appear to do what you want, think about your problem a
bit more and maybe you'll realise that you didn't fully understand your
Unsubscribe mailto:rrd-users-request at list.ee.ethz.ch?subject=unsubscribe
Help mailto:rrd-users-request at list.ee.ethz.ch?subject=help
More information about the rrd-users