Keeping Critical Systems Up 99.9% of the Time - Monitoring

(this article is part of a series)

The first step in keeping your critical systems online 99.9% of the time is to know when they're down.

If you're not monitoring your systems - at very best you know that things are working when you're using them. The other 90% of the time - you just have to hope things are running OK.

That potential customer who got a server error when accessing your website probably isn't going to try emailing you to let you know things are broken.

Even if you choose not to implement backup systems - monitoring is a good idea. More often than not - system failures are non-catastrophic. Getting things working again may simply be a matter of a phone-call to your hosting provider (although without a backup server, you're at their mercy to get things fixed). One way or the other - you have to know about the problem in order to fix it.

It took me a long time to find monitoring software I was happy with. When working for other companies I've used various "commercial grade" applications that worked great. The problem is - I'd never spend $1500 of my own money on one of them.

There are a number of monitoring products on the market from $75 to $150. From my experience - some of these are not too bad, but a lot of them just seem a little "junky." Monitoring software is a potential point of failure in any backup system. You should feel confident that whatever product you're using is of high quality.

One thing to keep in mind - server monitoring software should generally run on your computer as a "service." You don't want your monitoring to go offline if your computer gets rebooted, and then no one logs on. Also - running as a service reduces the chance of you accidentally closing the monitoring application.

I've been using Paessler's IPCheck for about 6 months now and am very happy. It seems to have "commercial grade" quality at a relatively low cost. "Full" licenses start at about $150 which can handle up to 25 different monitoring points. As always - take advantage of their free trial before spending any money.

The freeware version is limited to 5 "sensors" with a monitoring interval of 15 minutes or greater. That's not enough to insure 99.9% uptime - but it's a lot better than nothing. Considering the price of "free," anyone who cares if their website is up or not should give it a look.

Of course - your monitoring system is dependant on whatever computer it's running on and the infrastructure behind it (power, internet connectivity, etc).

I run my monitoring software on the main development system for the SpamButcher personal spam filter. It's certainly possible I could lose power or have a drive crash at the same moment my main server fails. If you're extra paranoid, you could try implementing redundant monitoring systems. The truth is that you have to draw the line someplace.

Your monitoring system should not be located on the same network as your primary web server. Part of monitoring your server is actually verifying it's available via the internet - not just the local network.

If you must locate your monitoring system on the same network, monitor two major external sites like Yahoo! and Google. If they both go offline at the same time - you can be pretty certain something is wrong with your connectivity, and action needs to be taken.

In the next articles I'll discuss what services you can and should monitor, and what you (or hopefully the monitoring software) should do in event of a failure.

Next article: DNS

Back