icinga2/doc/12-agent-based-checks.md

Additional Agent-based Checks

If the remote services are not directly accessible through the network, a local agent installation exposing the results to check queries can become handy.

SNMP

The SNMP daemon runs on the remote system and answers SNMP queries by plugin binaries. The Monitoring Plugins package ships the check_snmp plugin binary, but there are plenty of existing plugins for specific use cases already around, for example monitoring Cisco routers.

The following example uses the SNMP ITL CheckCommand and just overrides the snmp_oid custom attribute. A service is created for all hosts which have the snmp-community custom attribute.

apply Service "uptime" {
  import "generic-service"

  check_command = "snmp"
  vars.snmp_oid = "1.3.6.1.2.1.1.3.0"
  vars.snmp_miblist = "DISMAN-EVENT-MIB"

  assign where host.vars.snmp_community != ""
}

Additional SNMP plugins are available using the Manubulon SNMP Plugins.

If no snmp_miblist is specified the plugin will default to ALL. As the number of available MIB files on the system increases so will the load generated by this plugin if no MIB is specified. As such, it is recommended to always specify at least one MIB.

SSH

Calling a plugin using the SSH protocol to execute a plugin on the remote server fetching its return code and output. The by_ssh command object is part of the built-in templates and requires the check_by_ssh check plugin which is available in the Monitoring Plugins package.

object CheckCommand "by_ssh_swap" {
  import "by_ssh"

  vars.by_ssh_command = "/usr/lib/nagios/plugins/check_swap -w $by_ssh_swap_warn$ -c $by_ssh_swap_crit$"
  vars.by_ssh_swap_warn = "75%"
  vars.by_ssh_swap_crit = "50%"
}

object Service "swap" {
  import "generic-service"

  host_name = "remote-ssh-host"

  check_command = "by_ssh_swap"

  vars.by_ssh_logname = "icinga"
}

NRPE

NRPE runs as daemon on the remote client including the required plugins and command definitions. Icinga 2 calls the check_nrpe plugin binary in order to query the configured command on the remote client.

Note

The NRPE protocol is considered insecure and has multiple flaws in its design. Upstream is not willing to fix these issues.

In order to stay safe, please use the native Icinga 2 client instead.

The NRPE daemon uses its own configuration format in nrpe.cfg while check_nrpe can be embedded into the Icinga 2 CheckCommand configuration syntax.

You can use the check_nrpe plugin from the NRPE project to query the NRPE daemon. Icinga 2 provides the nrpe check command for this:

Example:

object Service "users" {
  import "generic-service"

  host_name = "remote-nrpe-host"

  check_command = "nrpe"
  vars.nrpe_command = "check_users"
}

nrpe.cfg:

command[check_users]=/usr/local/icinga/libexec/check_users -w 5 -c 10

If you are planning to pass arguments to NRPE using the -a command line parameter, make sure that your NRPE daemon has them supported and enabled.

Note

Enabling command arguments in NRPE is considered harmful and exposes a security risk allowing attackers to execute commands remotely. Details at seclists.org.

The plugin check command nrpe provides the nrpe_arguments custom attribute which expects either a single value or an array of values.

Example:

object Service "nrpe-disk-/" {
  import "generic-service"

  host_name = "remote-nrpe-host"

  check_command = "nrpe"
  vars.nrpe_command = "check_disk"
  vars.nrpe_arguments = [ "20%", "10%", "/" ]
}

Icinga 2 will execute the nrpe plugin like this:

/usr/lib/nagios/plugins/check_nrpe -H <remote-nrpe-host> -c 'check_disk' -a '20%' '10%' '/'

NRPE expects all additional arguments in an ordered fashion and interprets the first value as $ARG1$ macro, the second value as $ARG2$, and so on.

nrpe.cfg:

command[check_disk]=/usr/local/icinga/libexec/check_disk -w $ARG1$ -c $ARG2$ -p $ARG3$

Using the above example with nrpe_arguments the command executed by the NRPE daemon looks similar to that:

/usr/local/icinga/libexec/check_disk -w 20% -c 10% -p /

You can pass arguments in a similar manner to NSClient++ when using its NRPE supported check method.

NSClient++

NSClient++ works on both Windows and Linux platforms and is well known for its magnificent Windows support. There are alternatives like the WMI interface, but using NSClient++ will allow you to run local scripts similar to check plugins fetching the required output and performance counters.

You can use the check_nt plugin from the Monitoring Plugins project to query NSClient++. Icinga 2 provides the nscp check command for this:

Example:

object Service "disk" {
  import "generic-service"

  host_name = "remote-windows-host"

  check_command = "nscp"

  vars.nscp_variable = "USEDDISKSPACE"
  vars.nscp_params = "c"
  vars.nscp_warn = 70
  vars.nscp_crit = 80
}

For details on the NSClient++ configuration please refer to the official documentation.

NSCA-NG

NSCA-ng provides a client-server pair that allows the remote sender to push check results into the Icinga 2 ExternalCommandListener feature.

Note

This addon works in a similar fashion like the Icinga 1.x distributed model. If you are looking for a real distributed architecture with Icinga 2, scroll down.

Passive Check Results and SNMP Traps

SNMP Traps can be received and filtered by using SNMPTT and specific trap handlers passing the check results to Icinga 2.

Following the SNMPTT Format documentation and the Icinga external command syntax found here we can create generic services that can accommodate any number of hosts for a given scenario.

Simple SNMP Traps

A simple example might be monitoring host reboots indicated by an SNMP agent reset. Building the event to auto reset after dispatching a notification is important. Setup the manual check parameters to reset the event from an initial unhandled state or from a missed reset event.

Add a directive in snmptt.conf

EVENT coldStart .1.3.6.1.6.3.1.1.5.1 "Status Events" Normal
FORMAT Device reinitialized (coldStart)
EXEC echo "[$@] PROCESS_SERVICE_CHECK_RESULT;$A;Coldstart;2;The snmp agent has reinitialized." >> /var/run/icinga2/cmd/icinga2.cmd
SDESC
A coldStart trap signifies that the SNMPv2 entity, acting
in an agent role, is reinitializing itself and that its
configuration may have been altered.
EDESC

1. Define the EVENT as per your need.
2. Construct the EXEC statement with the service name matching your template applied to your n hosts. The host address inferred by SNMPTT will be the correlating factor. You can have snmptt provide host names or ip addresses to match your Icinga convention.

Add an EventCommand configuration object for the passive service auto reset event.

object EventCommand "coldstart-reset-event" {
  import "plugin-event-command"

  command = [ SysconfDir + "/icinga2/conf.d/custom/scripts/coldstart_reset_event.sh" ]

  arguments = {
    "-i" = "$service.state_id$"
    "-n" = "$host.name$"
    "-s" = "$service.name$"
  }
}

Create the coldstart_reset_event.sh shell script to pass the expanded variable data in. The $service.state_id$ is important in order to prevent an endless loop of event firing after the service has been reset.

#!/bin/bash

SERVICE_STATE_ID=""
HOST_NAME=""
SERVICE_NAME=""

show_help()
{
cat <<-EOF
	Usage: ${0##*/} [-h] -n HOST_NAME -s SERVICE_NAME
	Writes a coldstart reset event to the Icinga command pipe.

	  -h                  Display this help and exit.
	  -i SERVICE_STATE_ID The associated service state id.
	  -n HOST_NAME        The associated host name.
	  -s SERVICE_NAME     The associated service name.
EOF
}

while getopts "hi:n:s:" opt; do
    case "$opt" in
      h)
          show_help
          exit 0
          ;;
      i)
          SERVICE_STATE_ID=$OPTARG
          ;;
      n)
          HOST_NAME=$OPTARG
          ;;
      s)
          SERVICE_NAME=$OPTARG
          ;;
      '?')
          show_help
          exit 0
          ;;
      esac
done

if [ -z "$SERVICE_STATE_ID" ]; then
    show_help
    printf "\n  Error: -i required.\n"
    exit 1
fi

if [ -z "$HOST_NAME" ]; then
    show_help
    printf "\n  Error: -n required.\n"
    exit 1
fi

if [ -z "$SERVICE_NAME" ]; then
    show_help
    printf "\n  Error: -s required.\n"
    exit 1
fi

if [ "$SERVICE_STATE_ID" -gt 0 ]; then
    echo "[`date +%s`] PROCESS_SERVICE_CHECK_RESULT;$HOST_NAME;$SERVICE_NAME;0;Auto-reset (`date +"%m-%d-%Y %T"`)." >> /var/run/icinga2/cmd/icinga2.cmd
fi

Finally create the Service and assign it:

apply Service "Coldstart" {
  import "generic-service-custom"

  check_command         = "dummy"
  event_command         = "coldstart-reset-event"

  enable_notifications  = 1
  enable_active_checks  = 0
  enable_passive_checks = 1
  enable_flapping       = 0
  volatile              = 1
  enable_perfdata       = 0

  vars.dummy_state      = 0
  vars.dummy_text       = "Manual reset."

  vars.sla              = "24x7"

  assign where (host.vars.os == "Linux" || host.vars.os == "Windows")
}

Complex SNMP Traps

A more complex example might be passing dynamic data from a traps varbind list for a backup scenario where the backup software dispatches status updates. By utilizing active and passive checks, the older freshness concept can be leveraged.

By defining the active check as a hard failed state, a missed backup can be reported. As long as the most recent passive update has occurred, the active check is bypassed.

Add a directive in snmptt.conf

EVENT enterpriseSpecific <YOUR OID> "Status Events" Normal
FORMAT Enterprise specific trap
EXEC echo "[$@] PROCESS_SERVICE_CHECK_RESULT;$A;$1;$2;$3" >> /var/run/icinga2/cmd/icinga2.cmd
SDESC
An enterprise specific trap.
The varbinds in order denote the Icinga service name, state and text.
EDESC

1. Define the EVENT as per your need using your actual oid.
2. The service name, state and text are extracted from the first three varbinds. This has the advantage of accommodating an unlimited set of use cases.

Create a Service for the specific use case associated to the host. If the host matches and the first varbind value is Backup, SNMPTT will submit the corresponding passive update with the state and text from the second and third varbind:

object Service "Backup" {
  import "generic-service-custom"

  host_name             = "host.domain.com"
  check_command         = "dummy"

  enable_notifications  = 1
  enable_active_checks  = 1
  enable_passive_checks = 1
  enable_flapping       = 0
  volatile              = 1
  max_check_attempts    = 1
  check_interval        = 87000
  enable_perfdata       = 0

  vars.sla              = "24x7"
  vars.dummy_state      = 2
  vars.dummy_text       = "No passive check result received."
}