|Website occasionally unavailable for about 1 minute, but server not down - DNS or network issue?|
| 8:13 am on Jul 2, 2009 (gmt 0)|
We moved a few UK sites from a Canadian host to a US host in March 2008 and ever since we've occasionally been unable to access any sites on the server for about a minute, continually getting the "Oops this link appears broken" page.
When the problem occurs for me, other people in the office can still reach the page I can't get to.
We recently noticed this happening more and more, then on Tuesday 23rd June our traffic suddenly dropped by about 25% compared to normal. Our server monitoring reports show a sudden change in response times and transfer rates around 22nd/23rd June (response going from 0.8 seconds to 1.2, transfer rate going from 800kbps to 600kbps) and it has remained that way since.
I contacted my hosting company and they said a "a provider of ours suffered a fiber cut that has shifted how some bandwidth is routed into our network"........"traffic should normalize for you after the repair is completed, but I cannot guarantee it".
I'm wondering if this, the user specific unavailability of the site for short periods and our drop in traffic are all related?
Any advice, comments would be appreciated.
| 1:11 pm on Jul 2, 2009 (gmt 0)|
Does the error message give an error code or any indication of what specific error is being generated? It does sound like a routing error, you can try doing a traceroute to see if it times out before reaching your server.
If the routing error is affecting you, then it is surely randomly affecting other visitors too. The responsibility is firmly in the hands of your hosting company, if the situation persists and they are unwilling to help, then you should start to look elsewhere for a hosting solution.
| 1:43 pm on Jul 2, 2009 (gmt 0)|
There's no error message, it's the standard Internet Explorer page cannot be found page.
We raised this in December last year and managed to run a DNS report when it happened which revealed one of the hosts nameservers wasn't responding at that precise moment. Seconds later everything was fine again.
Our hosts made some adjustments to zone files etc. and things seemed to improve although the problem didn't go away. It now seems more noticeable again.
As you say, if I notice it, many of our potential visitors must do also, and probably go elsewhere. I'm not sure if this is something that could also impact on our search engine rankings as well, but it seems reasonable to assume robots encounter the issue occasionally as well.
Maybe the recent network issue has highlighted the underlying issue which sounds like it could be DNS related again?
| 3:29 pm on Jul 2, 2009 (gmt 0)|
Which version of IE are you using? IE8? The standard IE8 error page usually has a "More information" link which should give some hints. If it offers to diagnose connection problems and gives options such as "The website is temporarily unavailable" or "The Domain Name Server (DNS) is not reachable", then you probably have a DNS problem.
Are your nameservers provided by your hosting provider? How many are listed? Do you run bind or a similar DNS service directly on your own server, or are you using shared hosting?
| 4:21 pm on Jul 2, 2009 (gmt 0)|
I'm on IE8 but I get the Google "Oops, this link appears broken" message.
We have a dedicated server using our host's primary and secondary nameservers.
Just had the problem again actually. I quickly ran a DNS report and did a traceroute but all came up okay. I'm fairly confident it's something in our DNS set up - it's just a matter of catching it so I can prove it.
Also noticed in the traceroute details that the route is London, Amsterdam, Dusseldorf, Frankfurt, Chicago, host nameservers whereas before it went London, New York, Chicago, host. That's added 20 milliseconds and a few more places it could all fall down I suppose.
I'm pretty confident it's their nameservers but I can't get over the coincidence of our traffic drop on the same day they had the "fiber cut" incident with their network.
| 4:40 pm on Jul 2, 2009 (gmt 0)|
Are you using the Google toolbar or similar tool? When you get "Oops, this link appears broken", is it followed by any more precise description of the problem?
You should try disabling the toolbar's interception of server messages so you get the standard Windows or remote server error messages instead. See here:
| 5:06 pm on Jul 2, 2009 (gmt 0)|
Made the change suggested, thanks for that. Hopefully that will help shed some light on the issue.
| 7:47 am on Jul 7, 2009 (gmt 0)|
Just had my server monitoring reports for last week which show performance returned to almost normal on Tuesday/Wednesday last week and our traffic began recovering as well - traffic for Monday 6th July was almost identical to the Monday before the fiber cut issue.
I wonder if the fiber cut incident exagerated an underlying DNS issue and prevented even more visitors making it to the site.
The 'oops this link appears broken' issue has happened less since the fiber cut issue was fixed, but it's still happening so I have a feeling we're still losing visitors but at a much lower rate.
Time to hassle my hosts a bit more I think. I'm convinced this is an issue with their nameservers. The only time we've been able to prove anything it was there secondary nameserver that didn't respond.
| 12:43 am on Jul 9, 2009 (gmt 0)|
Often you get hosts offering two nameservers, often on the same network and sometimes being the same physical server - this leaves a single point of failure for DNS resolution.
Ideally, you should have at least three nameservers on different IP blocks, and geographically separated. There are a few companies offering DNS services (for a fee) and which are often much more robust than a hosting company's DNS setup.