EM Supportability Metrics - Internal - Threads

[David Richards]

Does anyone have any documentation regarding the "Threads" metric in the EM supportability metrics ?

It looks like a lot of data is being gathered, and we use Thread monitoring extensively to look after application servers, so I was wondering if any of them are usable to help monitor the EMs as well.

Thanks

[Anon Anon]

Hello Wily Community:

Any assistance here for Dave?

Thanks,
Mary

[David Richards]

I am still interested in this topic. Is there anybody from CA who might have some docs around what the Supportability metrics are telling us ?

Thanks

[Guenter DI Grossberger]

Hi Dave,

there's no public documentation about the supportability metrics. I don't know anything specific about the threads metrics apart from the obvious that you can guess from the metric names.

The most relevant supportability metrics are:

* all metrics shown in the overview typeview when selecting Custom Metric Agent (Virtual)|Enterprise Manager

* Connections: these show you how many metrics the EM is getting from how many agents and how many workstations are querying the data

* GC Heap: obviously showing the memory usage of the EM

* Health: estimated used vs free capacity of the EM regarding several important factors like CPU, heap or incoming data

* Tasks: Harvest and SmartStor Duration. "Harvest Duration is probably the most important metric in assessing a performance or capacity problem. This is the duration it takes to process all of the incoming agent connections and metrics. When this exceeds 3.5 seconds, very bad things are going to occur as all other activities are subordinate to the Harvest and this means less time for the remaining EM activities." SmartStor duration is obviously the time it takes to write all metrics to disk.

Unfortunately I don't have more information about the threads but it has not turned up as something very important or influencial in my three years at CA wokring with APM.

Best regards,
Guenter

[Legacy User]

Guenter,

Great response. Out of curiosity, what factors have the most impact on Harvest Duration, e.g. Disk I/O, CPU? In other words, if an APM team notices a large spike in Harvest Duration, where should a team look first to resolve or mitigate the issue?

Regards,
Jack

[Anon Anon]

I would first look a the number of Metrics coming in to see if there is any Metric explosion because of some improperly configured Agent profile file.
Next I would look at i/o on the Server where SmartStore is located (if Unix, iostat will do)
Next I would look at CPU utilization on the Server were Enterprise Manager is running - are you hitting the ceiling ?
Next I would look at the GC Heap of the Enterprise manager to see if I'm spending lot of time in Garbage collection.

Good luck.

[Anon Anon]

Very useful. Thanks Guenter !