App: Metricator for nmon

[damucka]

Hello,

I deployed the Metricator for nmon app and I like it.
There is however one functionality that I am missing, perhaps someone could advice if it is not there or I just cannot find it:
- it would be very helpful when one could label the monitored server and also group them. Quite often systems consist of several servers and I would like to see the KPIs per system / group, like e.g. the average CPU utilization of the group.

Could you advise?

Kind Regards,
Kamil

[guilmxm]

Hi @damucka

Please upgrade to the latest version from Splunk Base, version 1.0.23, this fixes the TOP stats issues in the wall of performance UI, and this provides better and correct filtering options in the Alerting center UI.

Do you still have issues with the frameID mapping ?

Guilhem

[guilmxm]

Hi @damucka

Sorry for late reply, being a bit busy right now.
The alerting dashboard you mention has free text form filters, unlike a dropdown or a multi dropdown these will not generate anything but allow you write something.

So these are due to allow manual filtering.

Regarding the frameID identifier, when a UI provides a dropdown selector or a multi selector, the list of the hosts that are provided should be only the hosts that are linked to that frameId.

Can you past paste a screenshot example like in the CPU_ALL user interface for example ?

I will come back with a few screenshot to demonstrate that.

Guilhem

[damucka]

Hello Guilhem,

Thank you.
Still I am not able to understand what is the functionality of this filter, to me it seems there is none.
Is there any way to filter only the alerts relevant to my landscape (FRAMEID) or specific server on the alerting dashboard?

Also I have a question regarding the detailed CPU statistics under "NMON Performance - Wall Of Performance":
- it seems to me, that the scale of the logical CPUs under "Deployment TOP 10 CPU processes usage" and "By host TOP 10 CPU processes usage" is false. It shows the usage of e.g. 20,000 cores ... we do not have so many. It is a bit strange. The % CPU usage looks correct.
The scale of the "Logical cores" under the NMON Analyzer for Linux --> Process, Kernel, I/O Statistics --> TOP, CPU Usage per logical core seems also to be correct.
Just the WALL OF Performance seems to be wrong for the logical CPUs.

Could you please advice?

Kind Regards,
Kamil

[guilmxm]

Hi @damucka

Right, thanks for leveraging the issue with this UI.
Indeed, the form filters are operating on a small portion of the charts, so it's more confusing than useful.
I will review this UI shortly.

Regarding the scale of the logical CPU cores between the Wall of performance UI and the Analyser / dedicated view, can you confirm that you are looking at the exact same hosts selection ?

I will review this as well.

Guilhem

[guilmxm]

Hello @damucka

This is a perfectly fair question, and feature that I understand completely.

So, in the context of the application, this is the purpose of the "frameID" identifier that is available in each user interface, and used widely across the reports and dashboards.

This is documented here:
https://www.octamis.com/metricator-docs/frameID_mapping.html

Basically you can define a value of frameID for a server X either on the raw metrics level at ingestion time, or at search time via a lookup mapping that is achieved with a KVstore collection and the associated lookup.

A user interface is provided to allow you doing the frameID change for a server, and you could as well use any other mechanism such as a search with a conditional statement or the lookup editor app.

By default and except for the IBM based systems (for which the frameID is automatically set to the ID of the frame that hosting the LPAR) the frameID will be equivalent to the host name value.
Just follow the doc or directly use the UI to associate the servers together using the frameID.

Hope this helps, feel free to let me know otherwise.

Guilhem

[damucka]

Hello Guilhem,

Thank you.
I used the configuration interface to set the new FRAMEIDs, but I am not able to figure out how I should define it in order to achieve the host grouping effect. My observations are:
- when I add the FRAMEID with one host only, then it is visible in the UI, that is fine. Perhaps I would wish myself that when I choose the FRAMEID having single host, then it automatically gets populated and I do not have to choose it manually
- however what I need is one FRAMEID having several hosts grouped in it, like my DB has several nodes. Then I would like to have e.g. FRAMEID named SHT_HDB with two hosts ls5749 and ls6258 attached to it. Then choosing the SHT_HDB in the UI, I would expect the corresponding hosts being transferred to the field "Host" and the aggregated numbers being displayed (CPU. Memory, etc.).
Or is the FRAMEID only the labeling of the single hosts without the functionality of grouping?

I tried to configure it in many ways using the interface, but I failed. If I try to configure more than one host, the FRAMEID gets saved, but does not show in the UIs then.
Is it possible to do what I want with FRAMEID?
If yes, could you please let me know what structure the FRAMEID definition should have in the configuration interface given the above example?

Kind Regards,
Kamil

[damucka]

Hello Guilhem,

I guess I got what I wanted, it works quite well with the "NMON WALL OF PERFORMANCE" functionality. When I set the same FRAMEID for group of servers that differ on the host name and serial number fields, then it works fine.

Another question:
- I noticed that when I enter the FRAMEID in the search criteria of the ALERT CENTER DASHBOARD it does not get populated to the panels - I still get the alerts for all configured hosts.
Am I misunderstanding something?

Kind Regards,
Kamil