[rrd-users] Time Shift on Trend Graph

Alex van den Bogaerdt alex at vandenbogaerdt.nl
Tue Jul 2 12:27:12 CEST 2013

----- Original Message ----- 
From: "Chris Mason" <chris at netnix.org>
To: "Alex van den Bogaerdt" <alex at vandenbogaerdt.nl>
Cc: <rrd-users at lists.oetiker.ch>
Sent: Tuesday, July 02, 2013 9:55 AM
Subject: Re: [rrd-users] Time Shift on Trend Graph

> Hi,
> Thanks for your response, but I have since stumbled across this post:
> http://oss.oetiker.ch/rrdtool/forum.en.html#nabble-td1514381
> I am trying to do the same thing as was discussed in this post (i.e. have
> the first function actually return the first point in the RRA with an
> actual data value).

You may have, but I see nothing there, and when I search for td1514381 I get 
only one post: yours.

> In simple terms, I am trying to make my graph grow from the left instead 
> of
> the right and I am looking for a <TIME> to pass "rrdtool graph -s <TIME>"
> to achieve this.

You have repeated yourself a couple of times now.  Each time that you try, 
you end up with nothing.

Computing <TIME> should be as simple as (rrdtool last)-(rrdtool first) and 
applying some corrections maybe.

What happens if you have <TIME> ? You just know how much data is supposed to 
be in the RRA, and how much data you are supposed to be able to graph, but 
it will still be past towards future, left to right.

> I am going to look at storing the create time of the RRD and comparing it
> with the response from first to see what is the latest value - however, I
> would still need to round my create time up to the next RRA interval
> otherwise I might have a bit of white space at the beginning.

Some more hints:

I don't think left growing is possible, unless maybe you could use a 
negative time scale.  I don't know, never tried it. It would probably result 
in a backwards graph anyway.

When you talk about Time Shift, I immediately think about the SHIFT 
function, which is described in 
It works similar to STACK, except that it works on the X-axis, not the 

If you are talking about the time shifting done by functions such as 
'predict' and 'trend', no actual shifting on the graph should happen.
But I do not know how this would fit in your problem description

I mentioned some docs to read, and suggested that you rephrase the problem. 
Unless you do, my guess is that nobody is going to be able to help.

I also guess that once you are actually breaking the problem down into 
smaller bits and pieces, and are trying to describe what goes wrong and 
where, you may be able to figure it out yourself.

More information about the rrd-users mailing list