Free Software

Back to resources.
Back to web site.

hostchecker.pl - Service Availability Monitor.

v.1.04 - 10/19/2003

• Overview
• How does it work?
• Downloading
• UNIX Installation
• Tips For Host Monitoring
• Feedback / Questions

Overview

Hostchecker.pl is a simple program intended to automatically query services on internet hosts or ping them to determine availability. It can be configured to query nearly any tcp service (http, dns, citrix, rtsp, etc.) so that you can actively monitor your server and receive alerts when a service (or host) becomes unavailable.

Hostchecker.pl is written in Perl and uses the Net::Ping library.

How does it work?

For example, let's say that you have two hosts: server1 and server2. Each machine is configured to run Hostchecker in their /etc/crontab file and scan the other machine for availability. They could be sitting next to each other or located on different networks. For example:

In the event that server1 goes down, (or even if one of the monitored services on server1 becomes unavailable), server2 will take notice when it is next run and send a brief e-mail to the address you've configured to receive alerts. This could be a pager, cell phone, etc.

If server1 continues to be down, server2 will send one further alert (to avoid bombarding you with alerts) and then keep checking silently until service becomes available again. Upon the resumption of availability, hostchecker will take note and send an alert.

Downloading

The perl script, sample configuration file, and installation instructions can be retrieved here:

UNIX Installation

1. Unpack the tar file and place the 'hostchecker.pl' file where you intend to keep the program. (i.e. '/root'). The 'hostchecker.cfg' file should be placed in /etc.

2. Edit the /etc/hostchecker.cfg file to suit your needs. The format is sectioned into Config, BaseLine, and Monitor as follows:

<Config>
directory=/tmp
notify=someaddress@pager.com
sendmail=/usr/sbin/sendmail
</Config>

Generally the '/tmp' directory is a good place for the program to put its temporary files. Notify should be set to a valid e-mail address. (The alerts are size-optimized with pagers in mind.)

<BaseLine>
www.cnn.com,http
www.news.com,http
www.aol.com,http
etc...
</BaseLine>

The BaseLine section contains a list of hosts that should respond to your query under normal circumstances. You should have at least 10 or 20 for a good sampling. If 90% of these hosts do NOT respond to a query, it is likely that your machine is down and the program will terminate. (We don't want to queue up any alerts because they'll all be delivered in a flood once the host goes back online.)

<Monitor>
chicago1,smtp
chicago1,http
chicago1,ping
chicago1,domain
chicago1,ftp
chicago1,rtsp

san-anton2,domain
san-anton2,pop3
san-anton2,imap
san-anton2,ftp

citrixserver.someone.com,ica

border-router,ping
gateway-router,ping

firewall.mynetwork.com,ping

</Monitor>

As you can see by the sample, you can monitor a broad range of services using the program. (The service name must be defined in the '/etc/services' file. See that file for more information about service names.) Note: 'ping' is an exception that is coded within the program to do a simple ICMP query.

The format is simple: hostname.something.com,service

Why monitor so many services? Won't a simple 'ping' entry be sufficient? ... Not quite. It is very possible that an application error could occur without the machine going down. For example: Apache could crash--without a 'http' monitor, you won't be alerted to this.

3. You can now run the program in 'debug' mode to test whether it is working or not by executing:

./hostchecker.pl debug

4. As long as everything looks ok, you're ready to add an '/etc/crontab' entry to run the program regularly. (The example runs the script every 5 minutes.)

*/5 * * * * root /root/hostchecker.pl

Tips For Host Monitoring

In general practice, it's bets to use a machine to monitor only the connectivity of remote servers and not the availability of their applications. In the event that the remote link fails, not only will you get a notification of the ping failing, but a notification on every other service you're monitoring as well. So--every time something goes funky with a link, you'll get a LOT of messages.

The best bet for monitoring a host's services is to do so locally and let remote machines worry about monitoring the hosts's connectivity. If the host goes down, it will find out through it's baseline status run and not run service status checks until it goes back up.

Feedback / Questions

This program was written very quickly to address a very old script that had ceased to perform properly. The old script required a separate instance with command line options for each host to be queried and relied on an external 'ping' call that took 10 seconds per query regardless of success. This program uses the Perl Net::Ping library and will run in a couple seconds if all is well.

Anyhow, if you find this program useful, please let us know.

E-Mail: contact@csma.biz.