On Tue, Oct 22, 2002 at 01:19:57PM -0400, DelRossi Donald wrote: > we are trying to migrate an application that currently runs on NT to > VM/Linux. This application a type of heartbeat monitor. It sends a request > to see if a server is available and if not reports back to the application > owners via email or whatever. The issue is that when a connection is > requested and fails Linux seems to try forever to make the connection. They > claim that it does not work this way on NT and want to scrape the project. > Since I am a proponent of running Linux on s/390 I would like to see if > there is a way to correct this from a Linux operating systems view.
It's kind of hard to answer this without a bit more information. Is this a TCP connection? If so, two possibilities immediately spring to mind: one is that the Linux box does not have DNS correctly set up, and so you're having to wait for a name query to time out. Do you mean "forever" or "a long time" ? The other one is that maybe you need to set TCP_NODELAY to *on* in your socket options when you create the socket. Anyway, I'm pretty close to certain that this isn't a Linux problem, it's a socket programming problem. The Linux socket API (essentially BSD sockets) is not quite the same as WinSock, although it's close. Maybe it's something as simple as not checking for the right return value to indicate the connection failed. > Any help would be appreciated. Adam
