Sum of loaned data does not display

[sjvrijn]

Describe the bug
I am trying to display a combined utilization value for some of the multi-gpu machines I monitor using this example from the documentation, but the placeholder/sum is not shown in my graphs. There are no connection issues, and no error in the graph log either.

I expect to be missing something in my configuration, but having gone through the documentation multiple times, I can't find what I'm missing or doing wrong, so any help would be much appreciated.

To Reproduce

• A server redstone.example.com with two NVIDIA gpus and the nvidia_gpu_[utilization] plugin

• /etc/munin/munin.conf: (trimmed away all irrelevant machines/categories)

cgitmpdir /var/tmp
includedir /etc/munin/conf.d

# Generate all graphs on cron for faster page loads
graph_strategy cron

# Configure cgi url for dynazoom
#cgiurl_graph /munin-cgi/munin-cgi-graph

# Generate all HTML periodically on cron
html_strategy cron


[Servers;]
        node_order GPU

[Servers;GPU;]
        node_order Summary

[Servers;GPU;Summary]
        update no

        # works as expected
        cpuload.graph_title GPU machines - CPU utilization
        cpuload.graph_args --base 1000 -l 0 -u 100 -r
        cpuload.graph_vlabel CPU
        cpuload.graph_scale no
        cpuload.graph_category system
        cpuload.graph_info Current CPU utilization
        cpuload.graph_order \
                redstone=Servers;GPU;redstone.example.com:cpu.idle \
        cpuload.redstone.cdef 100,redstone,32,/,-
        cpuload.redstone.draw LINE

        # does not display `redstone`
        gpu_util.graph_title GPU machines - GPU utilization
        gpu_util.graph_args --base 1000 -l 0 -u 100 -r
        gpu_util.graph_vlabel GPU
        gpu_util.graph_category system
        gpu_util.graph_order \
                test=Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                redstone \
        gpu_util.redstone.sum \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization1


[Servers;GPU;runite.example.com]
        address runite.example.com

Expected behavior
An entry redstone that tracks the sum of the two GPU utilization values, akin to the donald_disk example from http://guide.munin-monitoring.org/en/latest/example/graph/aggregation-combined.html

Screenshots & Logs
graphs:

/var/log/munin/munin-update.log:

2021/05/31 11:45:00 [INFO]: Starting munin-update
[...]
2021/05/31 11:45:09 [INFO] starting work in 3045 for redstone.example.com (redstone.example.com:4949).
[...]
2021/05/31 11:45:20 [INFO]: Munin-update finished for node Servers;GPU;redstone.example.com (11.10 sec)
[...]
2021/05/31 11:45:23 [INFO] Reaping Munin::Master::UpdateWorker<Servers;GPU;redstone.example.com>.  Exit value/signal: 0/0
[...]
2021/05/31 11:45:49 [INFO]: Munin-update finished (48.90 sec)

/var/log/munin/munin-graph.log:

2021/05/31 11:46:10 Starting munin-graph
2021/05/31 11:46:55 Munin-graph finished (44.50 sec)

(The 45 second duration is because my full config tracks ~30 machines instead of just redstone)

Desktop

• CentOS Linux release 7.9.2009 (Core)
• Munin Version 2.0.66 (Release 1.el7)

[sjvrijn]

I have found the issue: the placeholder redstone requires a label to be defined before it shows up in the graphs.
I.e. a line gpu_util.redstone.label redstone had to be added

[sumpfralle]

Just for my better understanding:

Graph is missing?

        gpu_util.graph_title GPU machines - GPU utilization
        gpu_util.graph_args --base 1000 -l 0 -u 100 -r
        gpu_util.graph_vlabel GPU
        gpu_util.graph_category system
        gpu_util.graph_order \
                test=Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                redstone \
        gpu_util.redstone.sum \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization1

Graph is visible?

(.label is added before .sum)

        gpu_util.graph_title GPU machines - GPU utilization
        gpu_util.graph_args --base 1000 -l 0 -u 100 -r
        gpu_util.graph_vlabel GPU
        gpu_util.graph_category system
        gpu_util.graph_order \
                test=Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                redstone \
        gpu_util.redstone.label foo
        gpu_util.redstone.sum \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization0 \
                Servers;GPU;redstone.example.com:nvidia_gpu_utilization.utilization1

Do I understand you correctly?
Is this related to .cdef vs. .sum (see your first example)?
Thank you!

[sjvrijn]

@sumpfralle Yes, with the gpu_util.redstone.label foo line added the entry becomes visible on the graph.
Difference is not due to cdef vs sum, but because any entry displayed in a graph needs two things: data and a label. When creating an entry using an initial placeholder name under graph_order, both still need to be specified.