HyperV-NSClient-PerfCounters

One script to automatically generate NSClient-compatible performance counters in Nagios, one to read them

Purpose

The final product of these scripts is a file with auto-generated Nagios/Icinga services. Those services will call a Python script that will contact the Hyper-V host to retrieve counter data.

Prerequisites

You need to bring:

A functional Nagios, Icinga, or compatible environment
check_nrpe on the above
A plug-in for the above that can process performance data, such as PNP4Nagios
A Python environment on the above
A functional Hyper-V host with one or more running virtual machines
Access to that host Hyper-V host with PowerShell
NSClient installed on that Hyper-V host, reachable by the Nagios/Icinga/compatible environment

Features of Python script

This section explains why you can't just use the normal performance counter reading tools and how the Python script solves them

If the target VM is off, no counter data will be generated (counter won't even exist); the script substitutes empty data
If the target VM is clustered, the script will chase down the owning node and retrieve counter data from it

How to use the PowerShell script

Read the help data for it.

How to configure NSClient

Make these changes to nsclient.ini:

[/settings/NRPE/server]
allow nasty characters = 1

[/modules]
CheckCounter = 1
CheckWMI = 1

Create the Nagios command

The automatically generated services list will attempt to use a Nagios command named "check-hypervmperf". How you implement that command depends on if you are using a secure or insecure channel to NSClient.

Insecure command

# Retrieves performance information for Hyper-V virtual machines
# HOSTADDRESS: always attach to the cluster or host that owns the virtual machine
# ARG1: the host or cluster that owns the virtual machine
# ARG2: the name of the virtual machine. for clustered VMs, the cluster resource must end in the same
# ARG3: the counter to retrieve. works with any counter on the host, but VMs are intended. ex: "\\Hyper-V Dynamic Memory VM(VMNAME)\\Physical Memory"
# ARG4: additional optional arguments to pass to check_hypervmperf
## -l # treat VM as clustered; find actual current host
## -W value # warn level for counter
## -C value # critical level for counter
## -m # when specified, values from -W and -C will be treated as lower bounds instead of upper (min instead of max)
define command{
   command_name check-hypervmperf
   command_line $USER1$/check_hypervmperf -H $ARG1$ -N $ARG2$ -c $ARG3$ $ARG4$
}

Secure command

Treat this as an example only. You will need to supply your own certificate paths and names.

# Retrieves performance information for Hyper-V virtual machines
# HOSTADDRESS: always attach to the cluster or host that owns the virtual machine
# ARG1: the host or cluster that owns the virtual machine
# ARG2: the name of the virtual machine. for clustered VMs, the cluster resource must end in the same
# ARG3: the counter to retrieve. works with any counter on the host, but VMs are intended. ex: "\\Hyper-V Dynamic Memory VM(VMNAME)\\Physical Memory"
# ARG4: additional optional arguments to pass to check_hypervmperf
## -l # treat VM as clustered; find actual current host
## -W value # warn level for counter
## -R value # critical level for counter
## -m # when specified, values from -W and -C will be treated as lower bounds instead of upper (min instead of max)
define command{
 command_name check-hypervmperf
 command_line $USER1$/check_hypervmperf -t 30 -p 5666 -A /var/ca/ca_cert.pem -C /var/certs/check_nrpe.pem -K /var/certs/check_nrpe.key -H $ARG1$ -N $ARG2$ -c $ARG3$ $ARG4$
}

Sample Templates

Configuration mismatches will cause perfectly logical output from this script to fail spectacularly. For instance, if you do not properly attach the service or host name to a valid parent, it might attach to something like "generic-host", causing all sorts of duplicate service warnings. Carefully think through and build out the template hierarchy to use for your VM performance counters before running this script. Connecting a Hyper-V host to its services can be done in multiple perfectly valid ways. Here is a suggestion:

## Virtual Machine object ##
define host{
   name hyperv-vm
   use perf-host,generic-host
   contact_groups admins
   check_command check-vm-null
   max_check_attempts 1
   register 0
}

define command{
   command_name check-vm-null
   command_line $USER1$/check_dummy 0
}

define servicegroup{
   servicegroup_name hyperv-vm-perf-services
   alias Hyper-V VM Performance Metrics
}

define service{
   name hyperv-vm-performance
   servicegroups hyperv-vm-perf-services
   use perf-service,generic-service
   contact_groups admins
   notification_period none
   register 0
}

Instruct the script to use the above host and service names as templates, and you should avoid collisions.

ejsiron/HyperV-NSClient-PerfCounters