node graph cleanups

[ecsv]

Fixes the down graph problem as described in issue #83 and the overlapping labels problem for nodes with many clients from issue #84

[jplitza]

Merged the part that corresponded to #84.

Did you test the part for #83 with high-resolution graphs as well? I applied it and notice that all the graphs now show the node as down during the current hour (example).

[ecsv]

I have the problem of nodes showing as down for the current time since a long time (before even thinking about the changes from #83). It seems to be related how the graphs are generated. It might be better to not graph from -xdays till now but instead -xdays until the time the gathering of data started which resulted in these graphs. But it could still end up with problems related to the aggregation.

But to your specific question: No, I am not using it together with graphs only showing the last two days.

Btw. the current code is following:

# this first one is most likely related to what you see
if !defined(upstate):
   upstate = 0
if upstate == 0:
   t = 1
if !defined(maxc):
    maxc = 1
maxc = max(1, maxc)
d = t * maxc

[ecsv]

I have now modified the code to assume that undefined upstate means that the node is running:

if !defined(upstate):
   upstate = 1
if upstate == 0:
   t = 1
if !defined(maxc):
    maxc = 1
maxc = max(1, maxc)
d = t * maxc

This is now a different interpretation but should avoid that the "down" region at the current time (most likely caused by the averaging taking place during the generation of the graph)

[ecsv]

Closing this because we are not using it anymore and there was no reaction anymore from upstream