Custom VM CPU latency % monitor

[dariusz_s]

Hello

I need some help with creating custom VM CPU latency % monitor in SCOM 2012. We use Veeam 6.5 management pack. I tried creating custom WMI monitor with 2 thresholds: one warning and one critical but it doesn't collect required data. I tried this:

Monitor target: WMware Virtual Machine
Parent monitor: Performance
WMI namespace: root/nworks
Query: select cpulatency from vmstats (this returns data when I run it using wbemtest), frequency 15 mins
Performance mapper: Object: VMGuest-cpu, Counter: cpulatencypct, Instance: _Total, Value: $Data/Property[@Name='cpulatency']$
Thresholds: low 3, high 5 as I want warning when above 3 % and critical when above 5%
I left helath and alerting as default.

What have I done wrong? Is monitor targeted at wrong object? Can you point me in the right direction, please?

regards

Dariusz

[sergey.g]

Hi Dariusz,

First of all, I wanted to make sure you know that we already have CPU Latency monitor in our Management Pack. It's called VM Compute Latency Analysis, it's checking both CPU latency and Memory latency and if any of them are above threshold - it will generate an alert. However this monitor has only one threshold for each metric and it generates warning if one metric is above threshold and error if both are above their thresholds. You can change thresholds with an override, you can set memory latency threshold to a very high value and this monitor will monitor only CPU latency.

With respect to your particular question, I'm afraid that it's not that easy with VM metrics. Your query will return all VMs monitored by a collector, so you either have to modify your query to return only a specific VM, or use filtering engine, which is not available in GUI, or use datasource from our Management Pack which is also unavailable in GUI.

Let me double-check few things and I'll provide more detailed explanation and instructions.
Thanks.

[dariusz_s]

Hi Sergey,

Thanks for your reply. Yes, I know about Compute Latency Analysis monitor in the MP. It was disabled in our environment for the reason I don't know but I was asked by our VMware team to create them custom monitor for only CPU latency % and with 2 thresholds - warning and critical.
I know it's not that easy with VM metrics because I've already tried couple of things myself but I'm stuck now.
I would be grateful for more detailed explanation and instructions. I'm going to wait for further info from you.

thanks

[sergey.g]

Hi Dariusz,

Thank you very much for your great question,

I'm attaching a sample Management Pack, it creates one monitor for VM. By default it's cpu latency, but you can change to whatever our collector publish to WMI (it depends on the Collector version).

Veeam.VMware.VM.Metric.Monitor.zip

I've created it with your threshold settings, so you can use it out of the box.

Monitor type usage desciption:

The motnitor is using Veeam-branded monitor type (Veeam.Virt.Extensions.VMware.CommonPerformance.3StateThresholdMonitorType), which takes WMI class (<ClassName>VMSTATS</ClassName>) and numsamples (<NumSamples>1</NumSamples>, how many samples to wait before triggering a monitor) as parameters. Since this monior type works for any metric in the configured class and for all instaces of the configured class, this monitor can be cooked-down (trust us this is reall y cool thing in terms of performance). Then you just have to take metric you want into the performance mapper (<Value>$Data/Property[@Name='cpuLatency']$</Value>, object, counter and instance names could be anything you want) and perform a filtering, so that each VM will have its own metric (<FilterExpression>...</FilterExpression>)

You can either specify thresholds right in the management pack (<Threshold1>5</Threshold1><Threshold2>10</Threshold2>), or override them in SCOM console. Alert text can be changed in SCOM console too.

Let me know if you have any questions regarding using our montior types or just about creating custom monitors for data gathered by Veeam MP.
Thanks.

[dariusz_s]

Hi Sergey,

Thanks a lot for your help on this

regards