Introduction

NMIS8 is a powerful network management system, but NMIS is not only a network management system: it performs Windows and Unix server monitoring out of the box if your server supports SNMP. NMIS8 works well with Windows servers and especially with the popular NET-SNMP SNMP daemon, which is widely used on Linux and other Unix systems.

The core of NMIS server management is the Host Resources MIB, which is standardised in RFC2790, both the NET-SNMP, Windows SNMP and VMware ESXi daemons implement the Host Resources MIB. But NMIS is not relying on SNMP exclusively; it has capabilities for performance and fault management of servers, as well as service (application) monitoring running on servers.

This document describes a few common approaches towards service monitoring with or without using SNMP.

NMIS8 collects and stores the following performance data from servers:

Processor load
Physical memory usage (not currently available from VMware ESXi)
Virtual memory usage (not currently available from VMware ESXi)
Disk usage (capacity and used)
Network interface usage

Service Monitoring

Checks Host Resources MIB for service status
Performs poll of TCP/UDP port
Runs send/expect scripts which return status (can perform full transaction over system and is highly extensible)
Run external programs which return status (and optionally extra values)

You can download NMIS from the Opmantek webpage.

Windows Server Support

Depends on Windows SNMP Service (often referred to as agentless server monitoring). NMIS will collect the following information from Windows Servers in addition to the already described information.

IP Utilisation
Server Memory Pages per Sec
CPU Interrupts
Memory Stats (commited and available bytes)
Number of Users

All of these information requires no Agent to be installed onto the Windows server other than the freely available Windows SNMP Service (part of the Windows installation and setup).

Unix Server Support

Depends on NET-SNMP daemon being installed and configured, usually installed by default on most Unix platforms. Instructions for installing the NET-SNMP daemon are included in the NMIS 8 Installation Guide.

Configuring NMIS to Monitor Services

An updated services file "Services.nmis" was distributed with later NMIS8 versions, and is attached to this page Services.nmis, this is compatible with earlier versions of NMIS8. This file includes definitions for using the script method to poll HTTP, as well as monitoring for a MySQL daemon, and various ports using NMAP connections. Once Services.nmis is copied to /usr/local/nmis8/conf, you should verify that you have the scripts folder in nmis8, this is included in the install directory. Running cp -r /usr/local/nmis8/install/scripts /usr/local/nmis8/conf will ensure the service definition files are available.

When a service "outage" is detected you will receive NMIS events and notifications if NMIS has been configured to do so.

Step 1

From the NMIS Menu, navigate to "System -> System Configuration -> Nodes (devices)"

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management 1 Edit.png (NMIS8 Service Management 1 Edit.png)

Step 2

From the list of nodes, select "edit" from the server you want to monitor services on.

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management 2 Edit.png (NMIS8 Service Management 2 Edit.png)

Step 3

From the services box, select the services you want to monitor, you can select several services.

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management 3 Edit.png (NMIS8 Service Management 3 Edit.png)

Step 4

When you are done, select "Edit and Update Node", after the next poll cycle runs, within about 5 minutes you will see that the services are now being monitored

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management Final.png (NMIS8 Service Management Final.png)

Step 5

You can also now select "Services" from the node menu bar and see a graph of all the monitored services.

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management Final 2.png (NMIS8 Service Management Final 2.png)

Starting with version 8.4.10G the service graphs also include the response time (as number and and graph) which is the elapsed time of the service test.

Defining new SNMP based Services to Monitor

This is how to define a new services based on using SNMP for the status of the service.

Step 1

Determine the name of the service process you want to monitor, for example to Monitor MongoDB on Windows, the process name is mongod.exe. As the MongoDB server state is reported as an attribute for the named process, you'll have to tell NMIS what process to look for.

Step 2

From the NMIS Menu, navigate to "System -> System Configuration -> Services"

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management - new service 1.png (NMIS8 Service Management - new service 1.png)

Step 3

Then select add or edit to add a new service or edit an existing service monitoring definition.

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management - new service 2.png (NMIS8 Service Management - new service 2.png)

Step 4

Give the service a (display) name, for Windows we call them service and for Unix we call them daemons. Insert the appropriate process name in the Service Name field, and select the Service Type service to tell NMIS that this is an SNMP-monitored service.

NMIS > Managing Servers and Services with NMIS8 > NMIS8 Service Management - new service 3.png (NMIS8 Service Management - new service 3.png)

Step 5

Click "Add" or "Edit" to save the new service and then you can assign this to a device as described above in "Configuring NMIS Services to Monitor".

Defining new TCP or UDP Ports to Monitor

Assuming you only want to check if a server listens for network connections on a particular TCP or UDP port (without actually exchanging application-level data with that service), then this is how to instruct NMIS to do that:

Select Service Type port
Give the appropriate port and protocol in the form of udp:123 (that would be NTP) or tcp:22 (that's SSH)
Ignore Service Name and all program-related settings

Please note that the open/reachable/closed status for UDP ports is unreliable due to the nature of the protocol; doing end-to-end service-specific tests is recommended for UDP.

Defining DNS Server Monitoring

NMIS also supports monitoring of DNS servers out of the box:

Service Type is dns.
Up to version 8.4.10G inclusive, the builtin dns monitor requires the node's Name/IP Address setting to hold the node's fully qualified domain name. NMIS then requests DNS info for this FQDN from the node that is marked for dns service monitoring. The Service Name is ignored.
After version 8.4.10G, you have to specify what is to be looked up as the Service Name. This can be either an IP address or a fully qualified domain name (host or domain). Again it is the node marked for dns monitoring that is asked for DNS info, but you've got control over what is requested.

Defining new Script based Services to Monitor

This is how to define a new services based on using a sapi script to connect to the socket and send some data, looking for the correct response.

Sapi scripts are extremely similar to expect scripts and documented at the end of /usr/local/nmis8/lib/sapi.pm.

Step 1

Determine the name of the service you want to monitor, for example to is the NMIS web application running.

Step 2 and Step 3

Same as above,

From the NMIS Menu, navigate to "System -> System Configuration -> Services"

Then select add or edit to add a new service or edit an existing service monitoring definition.

Step 4

Give the service a name, in this example nmis, and it is a script and define the port it will be here port 80.

NMIS > Managing Servers and Services with NMIS8 > NMIS Script Services.png

Step 5

Create a script which will get to the appropriate WEB Application URL, use the basic http service as a basis, the name of the service will be the name of the script

cp /usr/local/nmis8/conf/scripts/http /usr/local/nmis8/conf/scripts/nmis

Then edit that script and change the script to get your URL correctly, in this example the final nmis script looked like this:

send: HEAD /nmis8/ HTTP/1.0
send: 
expect: 200 OK

So this script will connect to port 80, send the "HEAD /nmis8/ HTTP/1.0" and a newline, the most basic if HTTP Headers, the response it would get would be:

HTTP/1.1 200 OK
Date: Tue, 01 Apr 2014 05:19:19 GMT
Server: Apache/2.2.15 (CentOS)
Last-Modified: Thu, 21 Mar 2013 05:14:21 GMT
ETag: "a233b-67c-4d8686950cd27"
Accept-Ranges: bytes
Content-Length: 1660
Vary: Accept-Encoding
Connection: close
Content-Type: text/html; charset=UTF-8

The expect would match the data "200 OK" and declare success. More complex requests could be made here. E.g. checking a database status using HTTP.

Monitoring Services with external programs

Starting with version 8.4.10G nmis can also run external programs to test a service status. To interoperate with NMIS such a program must conform to a few simple rules which are described below.

The steps for enabling this facility are mostly the same as for the other service types, except for the service definition of course. The dialog for Adding or Editing Services now shows some extra options:

NMIS > Managing Servers and Services with NMIS8 > service-dialog.png

Choosing the Service Type program activates the options for running external programs; they are ignored for all other Service Types.

Program Path must point to the program in question, and must be an absolute path (i.e. starting with /). This setting is clearly mandatory, and the program in question must be directly executable by the nmis user.
Program Args defines arguments that are to be given to this program when run. This is an optional setting.
Any tokens of the form node.something will be replaced by the corresponding property of this node; most likely you will want to use node.host for the node's dns name or network address, node.name for the logical node name, maybe even node.sysContact. To determine the available properties check the 'system' section of /usr/local/nmis8/var/yournode-node.nmis.
Max Program Runtime sets the upper limit (in seconds) for how long NMIS will wait for this program to complete.
If the program has not finished by that time then it's terminated forcibly and the service is marked 'down'. This setting is optional, but highly recommended - if you don't set a limit then a single uncooperative external program could delay all NMIS collect or update operations indefinitely!
Collect Program Output defines whether extra output by the program is to be collected and stored, or ignored.

The external program must follow these interfacing rules to work with NMIS:

Your external program must report the service status by returning an exit code between 0 and 100 inclusive.
0 means the services is down, any other value NMIS interprets as service is up. The service graphs do show the precise value, however.
Your external program can't read from stdin, and needs to finish its work and terminate as quickly as feasible; NMIS cannot proceed with further operations until the external program terminates.
Your external program may report back to NMIS by printing key=value pairs on stdout, one pair per line.
If the Collect Program Output option is enabled, then NMIS will collect these values and store them in the RRD database and the node file.
The key responsetime is special: any numeric value given for it will be used directly in the service and service response time graphs.

Monitoing services with an external program in this fashion is easy to implement, but limited in functionality; if you require access to all the advanced event management and alerting features of NMIS for a service then we recommend that you extend your snmpd for this service instead of using the external program facility.