Hi all,
my graphs are cut off on the right. I think I understand the root cause - there is simply no data (yet) in my InfluxDB for the time frame Garafana tries to display. But shouldn’t the “connected” setting for the “Null value” parameter take care of that? Any way to close that gap?
I’ll attach an example display including my configuration.
Thanks,
Patrick
No, the “connected” setting tells Grafana what to do about connecting together
(or not) two points which have a gap between them.
If you have a gap at the right hand side of a graph, then there’s a point to
the left of the gap, but there isn’t one to the right of it, therefore there
are not two points which can be “connected”.
Antony.
Understood. But is there a fix?
No; I’m saying that the request / expectation doesn’t make sense.
Suppose you take measurements of something every 10 minutes.
At 15:10 the reading is 3
At 15:20 the reading is 4
At 15:40 the reading is 7
That means the measurement for 15:30 is missing, in which case Grafana can:
a) join 4 to 7 with a straight line
b) show 0 for the missing value
c) show nothing for the missing value
Which of those is does is determined by what you’ve selected for the
“connected” option.
Now the time is 15:52, so the 15:50 value is clearly missing.
What is Grafana supposed to do?
a) it can’t join 7 to anything, because there’s nothing to join it to
b) it could show 0, but that might be wrong, if the 15:50 measurement appears
in the database sometime later
c) it could show nothing, which is what it actually does, on the basis that
there’s no information telling it what else would make sense.
Let me turn the question around: what would you expect Grafana to do at 15:52
for the missing 15:50 value?
Supplementary: what happens at 15:53 when the measurement turns up late?
Antony.
Well, what I would like it to do is extrapolate the last measurement to the right or move the graph only as far as there are measurments. Just to make the graph look nicer.
I am not quite sure if that is a reasonable expectation. I fully understand your reasoning. The problem of Grafana compared to simpler systems like Munin is that graphing and collecting metrics are losely coupled and asynchronous.
It only occured to me to ask that question because as soon as I through a selector or an aggregator like last() or mean() into my query, the lines extend all the way to the right. And I am still pondering why mean() is the default when adding a new query and over which period of time it actually averages - see my other post if you are interested in that conversation.
I am trying to really comprehend Influx and Grafana from scratch so I can design and run mission critical things for customers. Currently in the “messing around with my home NAS” phase. Gives me enough metrics to fiddle with for sure, since it is a rather high end system for home use, running 5 VMs plus 7 FreeBSD jails and so on …
Thanks for your help.
BTW: what does distinct() do that just field(value) doesn’t?
Kind regards,
Patrick
P.S. There’s a moving average function that could probably do the trick. Munin et al. essentially do the same with their measurements every 5 minutes.
OK, I think I found the “problem” - collectd seems to report the disk temperature only every 5 minutes. Whereas metrics like CPU or memory usage are collected way more frequently. And with a graph time frame of one hour, of course a 5 minute data point missing is noticeable.
