Originally posted by: Ross Whitehead
I have investigated the issue from early this evening, and believe the problem has been finally solved. It was the last piece of the puzzle... which made it all make sense.
So what was the "big secret" behind JRun? Was it the butler that did it?
After reading one of the other 12-page threads about it, from back in April, I had a couple of ideas.
Hypothetical: When memory gets loaded-down, CPU usage spikes, due to garbage-collection routines. Jason mentioned that the machine(s) in question are running Windows 2000, and it has been reported, and documented in MS's KB, that sometimes, under heavy load, W2K can "lose" network connections, mysteriously. Now, I'm not exactly sure what the architecture is on AT's servers, but if the "JRun connector" is actually over a network link between the front-end web server boxes and the back-end database server cluster nodes, then perhaps this could be an issue?
That was my working idea, until I saw your post that you think that it is now solved. My intellectual curiousity is dying to know exactly what it was, that was "bugging" these forums thus far. Did a mouse make a nest in the PSU fan on the server or something? Someone intentionally DDoS'ing the forums, perhaps with a "web scraper"? OS bugs? Bad NIC? Aliens trying to abduct the AT forum database, because they want to learn more about the "Grand Nagus"?