We are at the end of the road with this one !!!!
We have 20+ clients that we have upgraded to 5.0.5 or 5.0.6 over the last 2 years. Some of the upgrades involved new hardware as well.
The problem is that on an hourly basis, telnet sessions are "freezing" up. By whatever means (rebooting PC, closing emulation software, etc), the user starts a new telnet session, which gives them a new ttyp? number & everything appears to be fine.
However the ttyp? number they were using is still there with all the processes still running, but who, finger & whodo does not report the user as being logged in on that terminal. When you look at the process table using ps, you find that initial shell (ksh or sh) has a parent PID of 1 where current connections have a parent PID which is the PID of the telnet daemon (telnetd)
If these processes are not manually killed, they consume a license & eventually the server begins to slow down.
We have, on SCO's advice, changed network cards to 3C980C ones & we have even gone as far as changing one of our clients servers (at our own expense) to a Hewlett Packard server recommended to us by SCO Hardware Certification Team which made no difference.
I have a "gut feeling" that SCO 5.0.5 & above has a problem with faster processors (all affected machines are 800Mhz Pentium III or above) or is less tolerant of network transmission problems that older versions.
I am not asking for a solution (though it would be nice) What I need is some moral support & some evidence that I'm not going mad !! SCO have started to take the attitude that they are not interested & we should use OpenLinux instead !! With over 20 man-years work invested in Unix application software, we feel that is unacceptable !!!!
Thank in advance
Spencer Window (not a joke name)
spencer.window@eastmidlandcomputers.ltd.uk
We have 20+ clients that we have upgraded to 5.0.5 or 5.0.6 over the last 2 years. Some of the upgrades involved new hardware as well.
The problem is that on an hourly basis, telnet sessions are "freezing" up. By whatever means (rebooting PC, closing emulation software, etc), the user starts a new telnet session, which gives them a new ttyp? number & everything appears to be fine.
However the ttyp? number they were using is still there with all the processes still running, but who, finger & whodo does not report the user as being logged in on that terminal. When you look at the process table using ps, you find that initial shell (ksh or sh) has a parent PID of 1 where current connections have a parent PID which is the PID of the telnet daemon (telnetd)
If these processes are not manually killed, they consume a license & eventually the server begins to slow down.
We have, on SCO's advice, changed network cards to 3C980C ones & we have even gone as far as changing one of our clients servers (at our own expense) to a Hewlett Packard server recommended to us by SCO Hardware Certification Team which made no difference.
I have a "gut feeling" that SCO 5.0.5 & above has a problem with faster processors (all affected machines are 800Mhz Pentium III or above) or is less tolerant of network transmission problems that older versions.
I am not asking for a solution (though it would be nice) What I need is some moral support & some evidence that I'm not going mad !! SCO have started to take the attitude that they are not interested & we should use OpenLinux instead !! With over 20 man-years work invested in Unix application software, we feel that is unacceptable !!!!
Thank in advance
Spencer Window (not a joke name)
spencer.window@eastmidlandcomputers.ltd.uk