Help with the most simple xml file you ever heard of :-)

[nmonvisualizer]

Help with the most simple xml file you ever heard of :-)

[hpresnall]

Documentation bug - it should be --mt or --mf. I will fix that. However you don't need to multiplex by type unless you wan to get ResData for all the processes on each system.

Instead you just want to use -d to create a chart per hostname or -a to create one chart that has a line for each host. (If you want to use -a, you will need to add linesNamedBy="HOST" to the <linechart> element below.)

On your chart definition, you need to specify the type as ndsd like this:

<?xml version="1.1" encoding="UTF-8"?>
  <charts>
    <linechart name="NDSD ResData" shortName="NDSD">
      <yAxis label="NDSDResdata" />
      <data>
        <host os="linux" />
        <type name="ndsd" />
        <field name="ResData" />
      </data>
    </linechart>
</charts>

Note that this assumes you only have a single ndsd process on each host. If you have more than one process then the type named ndsd is actually the sum of the process data values for all the ndsd processes running.

[thsnielsen]

Thank you for your reply - that helped me progress a lot.
However (and sorry to pull off the arm when when you propose a hand) i am one step from perfect, and i would really like (another hand :-) to get there.
As you can see below, i have 3 ndsd processes one without PID, and two others with PID in brackets. I would the ResData for all three of them in the same chart.
Now if this gets complicated, just tell me, and i will stick with what i have it is already good, but if it is "simple" i would like to know.
Here my current xml file :
cat ndsdResData.xml

Here how my data looks like

[hpresnall]

Change linesNamedBy to TYPE and set the regex to ndsd.*;dot star - it's a regex, not a glob that always gets me too.

Note that the ndsd process without a number is the sum of all the ndsd processes. If you care about the total value, this may be all you need.

You may also just want a stack chart with all the processes on it (minus the overall ndsd process). To do that, add stack=true to the <linechart> element and change the <type> element to <type regex="java \((.*)\)" aliasByGroup="1" />. That type definition finds only java processes with parens and uses group 1 of the regex as the name on the chart.

[thsnielsen]

you rock !! Thanks a million, now i have exactly what i need, and a reasonable idea on howto change the xml for future needs.
Here for others what ended up doing :
java -jar NMONVisualizer_2015-10-21.jar com.ibm.nmon.ReportGenerator 20160112/* -a ndsdResData.xml

with

cat ndsdResData.xml

which gives :

exactly what i needed - THANK YOU :-)

[hpresnall]

FYI, you are naming by host but selecting multiple ndsd processes. What's being graphed is one of the ndsd processes on each host. Which one is somewhat random based on how the data is hashed internally.

If you want to graph all the processes on all the hosts change the linesNamedBy to HOST_TYPE. If you only want to graph the overall ResData, set the type element to just <type name="ndsd"> This will pick the aggregate process data.