Today, I was trying to check my domain's email, only to find that my server was down. I tried access my email and domain ten minutes later, and, thankfully, my site was back up. As far as I can tell, this is the first time I have experienced server downtime in over 6 months with my webhost -- although I say this because everytime I have tried to access my email and domain, I am successful at doing so.

I was wondering what can I do to accurately measure server downtime? And what is a reasonable amount of server downtime considering that my webhost provides a 99.9% uptime gaurantee?

Thank you,
Aeon

Well, there are monitoring services such as netwistle.com which offer some degree of notice. You have to keep in mind (if you use something like this), that the accuracy of the reports is only as good as the monitoring service itself, and well, sometimes these aren't 100% accurate.

Take netwhistle for example. It checks the server for free once an hour I believe. This is good except if the server is being rebooted, or apache is being restarted, etc., at the exact time netwhistle checks, then you'll get a server down report (and I believe it doesn't check again for another hour. So the downtime that it reported might have actually been 3 minutes.

You are allowed 1.44 minutes of downtime per day (99.9%).

I was, for some reason, under the impression that the uptime guarantee offered by web hosting companies in was a cumulative average, as opposed to a day-to-day guarantee. After all, someone can upload a script or a program, register a DLL, whatever, that can wipe out a server for half an hour, which isn't necessarily the web host's fault. And if they're not down at all for the rest of the month, that means they were down only 30 minutes, as opposed to the 43.2 they would be allowed for a 30-day month. Or again, am I totally off the mark in my thinking?

Every host has their own policy as to how they determine their uptime guarantee. I've read posts of people complaining that the uptime guarantee was network wide - not even pertaining specifically to the individual server that the person's site was on. In other words, the server could be off for 2 days straight, but so long as the overall average of the network (and all the servers) was above whatever%, then they claimed they were up. You really have to read what the uptime guarantee offers.

I also think people mistake what 99.9% uptime guarantees means. A host can refund your money for the downtime, but they surely can't guarantee the server will be up.

We use enotify.com

So far, so good

You know, I've been tinkering with writing a small utility you could run from your PC that would check a test page on your server at regular intervals -- minutes, hours, whatever.

The downside is, it would mean leaving your desktop rolling 24/7...

Ya think there'd be much interest in such a thing?

How about running a cron job every whatever interval you want to check, be it an hour, 30 minutes, 15 minutes etc... and append the current time to a log file. It should be very easy to implement if you are allowed to run cron jobs on the server.

If you are interested let me know.

Actually, I've done that... but what it tells you is if the box is running -- not if your site is live to the wire. Many (most?) hosts won't allow cron jobs, too.

I use server rat and it's pretty good but their software apprears to be quirkly and if you do go down even after they tell you you're back up you sometimes keep getting messages telling you it's up for a few hours... but then do check ever 15 minutes for free so who's to complain?

I was thinking along the lines of something that actually polled a (very small bandwidth) test page at intervals, direct from your own local system.

If you do this rightly, with a direct "get" of the page, it would even grab things like httpd failures -- a very common problem that none, if any, of the free externals will catch.

Nothing proves a server's up better than, well, going and getting an actual page off it...

Originally posted by etLux
If you do this rightly, with a direct "get" of the page, it would even grab things like httpd failures -- a very common problem that none, if any, of the free externals will catch.

Uptime http://uptime.arsdigita.com/uptime/about.tcl is a free service that gets a page from your site (a file with the word "success") via the web server every 15 minutes or so, seems to work OK.

Should be easy to write something similar to run locally or on a second server (it probably already exists).

Rob

Originally posted by etLux
You know, I've been tinkering with writing a small utility you could run from your PC that would check a test page on your server at regular intervals -- minutes, hours, whatever.

The downside is, it would mean leaving your desktop rolling 24/7...

Ya think there'd be much interest in such a thing?



I'd find that very useful.......since my desktop runs 24/7 online, and have 3 screens to put it on.