4.1.0 HP PolyServe Matrix Server administration guide (T5392-96073, October 2010)

Following are some examples of these strategies.
Device monitor or service monitor?
When deciding whether to monitor your custom application with a service monitor
or a device monitor, first determine whether the application provides a service over
the network directly to the outside world. If so, you should consider using a service
monitor. Because a service monitor is associated with a virtual host, it represents an
IP address provided to the outside world to access your services.
For example, one common use of Matrix Server is to increase the availability of Web
servers. The availability of the HTTP service is best modeled by a service monitor
associated with the virtual host to which outside users connect. Matrix Server provides
a built-in HTTP service monitor for this use. The other built-in service monitors offer
more examples of services which are best modeled by service monitors.
Alternatively, if you want to monitor something associated with a physical server,
you should consider using a device monitor. Monitoring the health of a local disk
drive, for example, is the job of a device monitor. The built-in DISK device monitor
is available for this purpose.
Similarly, if you have a Web server that communicates with a back-end database,
you might want to devise your own user-defined device monitor to determine whether
each server in the cluster has a good network connection to the database server.
Built-in monitor or user-defined monitor?
To decide whether to use a built-in monitor or a user-defined monitor, first determine
whether a built-in monitor is available for the service you want to monitor and then
consider the degree of content verification that you need. If you are monitoring a
custom application for which a built-in monitor is not available but the application
provides network services through certain ports, you may consider it sufficient to use
built-in TCP monitors on each of those ports.
However, if you want to verify that the application is still accepting connections to
the port and is also producing reasonable output, you should use a user-defined
service monitor. You will need to write a probe script that connects to the port and
then tests how the script responds to various commands.
Advanced monitor topics238