Ticket #174 (new defect)

Description

when the ferret drb server has too many requests going to it, especially large writes of long texts, the mongrels making the requests occasionally go into close_wait waiting for the drb server to finish, opening dozens or even hundreds of close_wait connections until the mongrel has to be outright restarted/killed. if the mongrel is not killed, it can keep opening close_wait connections until it is using all the open sockets and bring the website down. a large series of writes and reads to the ferret drb server brings out this behavior for us. we have a script running that has to kill a close_wait mongrel once every few hours for this very reason. by putting the mongrel into debug mode (killall -USR1 mongrel_rails) one can see the mongrel go haywire after making requests to the drb server, I believe.

yossarian1 gmail