For a few weeks now we've been experiencing consistent connectivity problems to SeqAnswers from our institute. I've been chasing this up with our networking people who say that they can't see any reason for the problem at our end so I'd like to try to track this from the SeqAnswers side as well. All page loads are currently taking around 30 seconds, but generate no errors and the site works normally (if very slowly).
As far as we can tell it's only SeqAnswers which is affected - we've not seen similar behaviour on other sites, and the problem seems to be related to our site, in that if I go to the site from my home connection then it's back to normal levels of responsiveness. Unfortunately the problem is bad enough that several people here have basically stopped using the site from work, which is a shame as they were making heavy use of it before.
I've tried to collect some useful diagnostic information to try to localise where the problem is happening, and have stuck some data up on our website.
We've tried running a traceroute to seqanswers but it looks like a server somewhere close to your site swallows the UDP packets so we can't get a full trace - but it looks quick enough up to that point.
I've also done a wireshark packet capture of a simple GET request for your index page. Looking at the results (which you'll need to view in wireshark) it looks like it takes 29 seconds to get an ACK response to our initial GET request which really does make it look like it's the seqanswers end which is taking the extra time - though I'm guessing our setup might be related.
If I were to take a wild guess I'd note that this slowdown happened at about the time we make some big DNS changes on our site. Is it possible that seqanswers is doing a reverse lookup on our IPs when we access the site, and that this is timing out each time we make a request?
I'm happy to spend some time debugging this as I'd like to get our users back on the site - any help we can get from the seqanswers end would be much appreciated.
Thanks
Simon.
As far as we can tell it's only SeqAnswers which is affected - we've not seen similar behaviour on other sites, and the problem seems to be related to our site, in that if I go to the site from my home connection then it's back to normal levels of responsiveness. Unfortunately the problem is bad enough that several people here have basically stopped using the site from work, which is a shame as they were making heavy use of it before.
I've tried to collect some useful diagnostic information to try to localise where the problem is happening, and have stuck some data up on our website.
We've tried running a traceroute to seqanswers but it looks like a server somewhere close to your site swallows the UDP packets so we can't get a full trace - but it looks quick enough up to that point.
I've also done a wireshark packet capture of a simple GET request for your index page. Looking at the results (which you'll need to view in wireshark) it looks like it takes 29 seconds to get an ACK response to our initial GET request which really does make it look like it's the seqanswers end which is taking the extra time - though I'm guessing our setup might be related.
If I were to take a wild guess I'd note that this slowdown happened at about the time we make some big DNS changes on our site. Is it possible that seqanswers is doing a reverse lookup on our IPs when we access the site, and that this is timing out each time we make a request?
I'm happy to spend some time debugging this as I'd like to get our users back on the site - any help we can get from the seqanswers end would be much appreciated.
Thanks
Simon.
Comment