Re: 4GL "Prompt" and zombies

In article <jlumbleyCpLwp5.2FK@netcom.com>, Joe Lumbley <jlumbley@netcom.com> wrote: >We have a 4GL (4GL version 4.11 and 5.01.UD4 on HP9000's) application that >is being accessed over telnet from a PC over the network. Occasionally >(we think the #*@# users are rebooting the PC), the 4GL program >becomes a zombie (top shows a ? as it's owning terminal) and the 4GL program >sucks up 95-100% of the CPU time. All the code is doing is waiting for the >user to enter data via a prompt command. > >Does anyone have any experience with this? On my system it happens all the time, though not necessarily with PROMPT. We run a dial-up BBS through which people access Informix applications -- some of our users have a bad habit of hanging up the phone rather than logging out. This, and some other circumstances, tends to leave an Informix zombie running that trashes the CPU cycles. Changing our Unix traps doesn't do anything ... we have had to have cron run a process every half hour thats looks for Informix zombies, checks whether they are holding a latch or in critical section, and if not kill -9's them (since we've never been able to make tbmode -z do anything, and normal kills don't work on the zombies.)