ServersCheck Logo

Thursday, July 31, 2008

Monitoring a website

A question we often get is: how do I monitor a website.

One has first to understand the different layers involved in a website and to what info he has access.

1. Hardware level:
A simple PING check will reveal if a server is alive or not. Monitor the response times as slower response times might indicate an error

2. OS level:
if you can then you can use the WindowsHealth or LinuxHealth check types to monitor the general health of the host running the site. This may help you to avoid situations like full dis, low memory resulting in a website failure.

3. Web server level:
A website runs on a machine, that hosts an Operating System and that has a web server. The site itself can be static or an application running as a layer on top of the web server.

A web server is typically monitored by using the TCP check to see if the server responds on the defined port (by default 80 or 443 for HTTPS)

4. Web application level:
The fastest check is to perform a HTTP HEADER check that verifies the header data sent back for a specific URL

The best option though is to use the HTTP check. It will load the given page and then scan the content to see if a predefined string is returned. If not, then an error is triggered.

You can also monitor the page download time; this is the time required to download the HTML - excluding images and any other embedded content.

Depending on your level access to your web site and server, you can define multiple layers of monitoring.

With ServersCheck you will know when your site fails before your customer(s) start calling you that there is an issue.

Monday, July 14, 2008

Email as best alerting option?

While getting through our Web Analytics reports, one thing that surprised me is that most people seem to be interested into a server monitoring software or environmental sensors with email as alerting capability.

Monitoring is not only detecting when something goes wrong but also making sure that you receive the alert when something like that happens.

Almost any competitive product on the market supports email alerting as it is the most basic way of alerting.

Unfortunately most customers are happy with this most basic way of alerting without realizing the potential risk of it. Some go a step further by having the email forwarded as a text message or on their PDA / Advanced Cell Phones.

Bottom line remains the same though: email requires network connectivity.

If your network is down, then your email alerts will not help you much. You will not get alerted and know that something serious is happening.

Therefore we stress on the fact that you should consider using SMS alerting using a GSM modem. I want to stress on "GSM Modem". As in the ServersCheck's software, quite a few other players offer SMS capabilities through gateways. ServersCheck uses ClickATell to date as one SMS alerting option. If you use SMS alerting through a gateway, the same critic as for email alerting applies: you will not get any SMS message if your network is down.

GSM modems are modems that are attached to a host computer (in our case the system running our software). They require a SIM card and behave like a cell phone. This means that through these GSM modems the software will send out the text message using the wireless phone network to deliver the text message. Even when your network is down you will get the alert message.

We want to stress on the importance of GSM modems as being one of the best ways of alerting (next to email - one does not replace the other). To emphasise this importance, we have decided to run this month a promotion whereby you get in North America & Europe a free FALCOM SAMBA USB GSM Modem for the purchase with a software license from ServersCheck (Business 500 or Premium Edition).

Monday, July 7, 2008

Impact of temperature on systems

Customers have reported to us server racks having caught fire due to an airconditioning failure resulting in an overheat in the racks.

Fire is not something we hear quite often as a result of temperature increase and the lack of temperature monitoring. Fortunately most servers nowadays have built-in security shutting down equipment when temperature is too high.

One of the issues people least are familiar with, is the effect of too high temperatures on CPU calculations. I am not talking in this article about processors melting down - yes that happens more than you think and way before systems catch fire.

Let's return to calculation errors.

A CPU is made of transitors and when those are becoming hot then they tend to leak current. This leakage causes calculation errors having impact on applications stability. Too many errors can cause a system crash.

The frustrating part of this is that initially errors go unnoticed. Only when applications start to fail, server and network monitoring software deployed will trigger alerts. If you want to act before it is too late, then temperature monitoring should be part of your overall monitoring strategy as it has a direct impact on applications & systems availability.

Maximum operating temperature of a CPU is often set around 40°C (104 F). Going above that 40° operating temperature and you are risking big issues rather sooner than later.