Aggregate Graph Plugin Issues with zeros

[mivok]

Hi. We currently have two firewalls, a primary and a backup, that we monitor
bandwith usage on, and have been using the aggregate graph plugin to create an
combined bandwidth graph showing the total bandwidth going across the two
firewalls.

The problem we are having is that most of the time, one of the two firewalls
(the backup) has no traffic going through it at all, showing a 0 on the graph.
This appears to be causing issues with stacking, where the positive value
(outbound in on this graph) is being stacked on top of a negative value rather
than on top of the X axis.

The way we have worked around this so far is to change the order of items in
the aggregate graph, grouping the inbound/outbound items together, making sure
there is an AREA item, followed by one or more STACK items, then another AREA,
followed by one of more STACK items. The original order was several AREA items
together, followed by the STACK items underneath.

This workaround has worked well so far, but we now have some extra machines we
want to include on the aggregate graph, and it is incredibly tedious to
re-order the graph items manually. Is there a way to have the aggregate graph
plugin automatically group similar items together, or is there another
solution we could use to prevent the incorrect stacking behavior?

I have attached several screenshots below, showing the aggregate graph item
order before and after we corrected it, as well as a zoomed in graph,
highlighting the issue we observed.

Thanks in advance,

Mark

[gandalf]

Which version of aggregate is this? I fear, that you've found a bug.
R.

[mivok]

I'm using version 0.71 with cacti 0.8.7b (plugin architecture version 2.1). I was using an earlier version of the plugin originally and tried upgrading to see if it fixed the issue.

[gandalf]

Ok, that's current. So will will have to investigate. But this will take time.
R.

[mivok]

Thanks, I appreciate it.

[gandalf]

On a closer look I see that this is a know design issue. I did not yet make up my mind how to handle that. So lets start doing it now, finally:

Due to the current methodology, aggregate simply adds one "host's graph items" after the other to the aggregate graph.
If the "base graph" already has a stack, this will yield unwanted results. Currently, the type (AREA, STACK) is computed correctly, but the order has to be changed after adding all graphs.
So I have to make an assumption, which order would be the correct one. I suppose, that the best approach would be to take the order from the "base graph", e.g. "input" first, then "Output". The resulting sort order would then be
- INPUT
-- "graphs" by listed sequence, first of them as AREA, rest as STACK
- OUTPUT
-- "graphs" by listed sequence, first of them as AREA, rest as STACK

While for that sort of graphs it should be mandatory, I'm thinking about making this an optional input parm to all aggregates. Sth like a dropdown:
- order by graph template item, graph
- order by graph, graph template item (this is the current approach)

R.

[gandalf]

This issue has been fixed in V0.73
R.