Re: Orphaned sql turbo processes.

Jonathan Leffler (johnl@informix.com) wrote: : >From: shyam@minerva.robadome.com (Shyam Davuburu) : >Subject: Orphaned sql turbo processes. : >Date: 13 Sep 1994 18:40:55 GMT : >X-Informix-List-Id: <news.8651> : Please remember to use newlines... : } I recently attend an Informix class on system administration, the : }class instructor claims when we C-ESQL code we need have a signal handler : }which do a sqlexit(). He claims that if the parent process gets killed : }then there is a chance that the sqlturbo will be hanging around. But I : }don't think so; sqlturbo use a pipe to communicate with the parent : }process, so the turbo process has to receive a SIGPIPE at one time or the : }other. Even if it ignores SIGPIPE it has to get an error when it tries to : }read the pipe, because the other end of the pipe is closed. Has anyone : }had the orphaned sqlturbo processes hanging around? Is there any logical : }reasoning why the sqltubo hanging around? I have not seen this problem, : }but informix guys are suggesting to us that we should have signal handlers : }in all our process, doing an sqlexit() call. : }Is it necessary to have a Signal handler doing this? I really appreciate : }your help on this. : The only advantage of having a signal handler around is that if the : sqlturbo is 20 minutes into a 2 hour query, the signal handler will stop : the sqlturbo before the query completes. On the other hand, if your system : is well behaved and queries don't ever take 2 hours, then there is : negligible benefit. As you say, if sqlturbo is writing to a process which : dies, then it gets SIGPIPE. On the other hand, and if it is waiting to : read, it gets an EOF indication (0 bytes read -- not an error, per se), and : it should go away fairly quickly. I use a Data General w/ Informix Online-5.00. When we have an orphaned sqlturbo process, we can usually kill it with a kill -15 command. I can't remember what this signal is, but it works well. Partha Dutta pdutta@netcom.com