Re: Killing Informix terminal sessions

Rule one --- NEVER, NEVER, NEVER (did I say NEVER) kill a sqlturbo process. Well, sort of never. If you kill a sqlturbo turbo process that is in the middle of a recovery, you *will* crash the database. Not might, *Will*. To demonstrate this, start a long transaction, then kill -15 the parent. Then immiediatly kill -9 the sqlturbo. Surprise!!! tbinit goes down. The only *correct* way, per Informix, to kill a process is to kill the parent process (ksh, nfglgo, or the actual program) with a kill -15 and WAIT!!!!!! Of course, that means you can't use 'trap "<process>" 15' in your UNIX system. We do use a killidle script ourselves, and it does use a kill -15. However, we occasionally see runaway disconnected processes also. ------------------------------------------------------------------------ In article <1995Nov17.121825.1@sysg.edsaus.com>, lnauhub.cz4dvt@eds.com (Malcolm Matheson) wrote: > > * HP-UX A.09.04 - running Informix-Online Version 5.05.UC2 > > I am considering the use of the public domain IDLED package (written > by Michael Crider) to terminate idle terminal sessions, but am > concerned about the possible side-effects on database integrity if > it terminates users running an application based on Informix. I've > looked at the Informix FAQ but couldn't get a clear answer. > > The IDLED package uses SIGKILL to terminate the processes. If this > is not safe for the Informix environment, what's the correct way to > terminate a specific user's processes? > > Any advice will be appreciated. > > Mal. > >=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= >| Malcolm J.H. Matheson | Voice: +613-9647-1693 | >| Systems Manager | Fax: +613-9647-1100 | >| EDS (Australia) Pty. Ltd. | E-mail: lnauhub.cz4dvt@eds.com | >=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=