Since launching 101bell.com, I’ve been plagued by a frustrating problem with outages. The server wouldn’t be down, but the outside world wouldn’t be able to get to the website. Then I launched the Phoenix Area Real Estate Blog, and the problem didn’t get worse, but the impact of the problem became more concerning for me.
What was frustrating, is that outages always seemed to occur on Saturday or Sunday, and there didn’t seem to be a cause. Last Saturday it again seemed to go “down” (again, the server was up and running, it just didn’t seem to respond), and I became determined to figure out what was wrong. Working from home, I logged into 101bell.net (which is a Windows Server), and from there logged into the server for 101bell.com. 101bell.com’s server is named azul, and it’s a Linux server.
I use the latest Debian release, which made installation a snap, and has made finding new software as well as updating existing software on the server equally easy. Easy for a Linux box, that is. It still takes some “skills”. And whatever was causing this outage seemed most likely to be a configuration issue with Azul.
So I looked closely at the basic network configuration. Azul has 2 network cards. One that sees the outside world (aka the internet) and one that sees the local network (intranet). I noticed that one of Azul’s DNS servers was 101bell.net. Although not a horrible setting, it’s not what I wanted as my T1 provider has perfectly good DNS servers that I’d prefer to use. I also noticed that when I did a traceroute to trace the IP traffic from Azul to an internet server (like google.com), that the route always went through 101bell.net.
That was certainly not right. Why was Azul not using the other network port and going directly out to the internet? Finally, I looked at my /etc/network/interfaces file and closely examined the configuration there. And sure enough, there it was as the first route to try: 101bell.net as a gateway. So I removed this entry from /etc/network/interfaces, and whala! the server really seems to be responding better now.
Although I was never really ever able to explain why it worked most of the time, but not all of the time, it has now worked all of the time for the past 48 hours. Google keeps track of unreachable links, and as of right now all links for all of the websites are up. I also noticed that 101bell.net has fewer network errors in it’s logs. So while my configuration still might not be perfect, it seems to be better.
Of course, I don’t expect anyone reading this to understand I’ve just written, but hopefully you’ll appreciate what a hassle hosting your own server can be. There’s a lot of times when I think, “For $50/month Godaddy.com would be happy to host my server for me.” One day I hope that my blogs will be so busy that I’ll need to move to Godaddy, just to keep up with the traffic. But for now, I’m building up my experience the hard way, so that when and if I need to move to Godaddy, I’ll at least know what my requirements are, which I couldn’t even say 6 months ago.