Re: "tbmode -z" doesn't kill - or even wound!

In article <435178918wnr@northop.demon.co.uk>, Eric Warner <Eric@northop.demon.co.uk> says: > >Hi, we've recently inherited an Informix On Line 4.1 system on a SCO >server and I've been elected DBA - on the basis that I knew what QWERTY >meant. > >I've had some formal training where I was told "Don't ever use kill -9 >to kill an Informix process - use tbmode -z instead." "EVER!" > >I needed to kill off a telnet login today for the first time, so <open >book> tbstat -u to get the pid then tbmode -z pid to kill it. Not. > >Hmmm, ps -ef then try to tbmode -z the parent process. Nope. > >Hmmmmmm, called The Bearded One who tutted for a while and then finally >killed it with "kill". > >So the question to you experienced folk out there is "Is tbmode -z just >one of those Good Things that nobody actually does in real life or was >I just unlucky?" > >Hopefully, > >Eric. > >=============================================================== >Eric Warner Northop, Clwyd, North Wales, U.K. > >Disclaimer: I disclaim everything. >=============================================================== > > Eric, I just ran across something very similar just today. I am very new to UNIX and INFORMIX also. When I did a "tbstat -u" I had two processes that had some problems. 1st process - In progress, waiting for latch, and in write holding up a checkpoint. 2nd process - In progress, waiting for checkpoint. I did a "tbmode -z pid" on the first process (which INFORMIX documentation says never to do or you risk a full meltdown!!!), but what occurred was that the latch cleared and the process continued in normal fashion, and then the second process finished after a checkpoint was finally executed (after approximately 1 1/2 hours). This seems very strange to me. Both processes were updating indexes in different tables, but this HOLD situation can take up to 48 hours -- with no activity in online.log and no reads and writes occurring. All we are attempting to do is to an UPDATE with an embedded sub-query. So, in my limited experience a "tbmode -z" does not do exactly what INFORMIX documentation says it should. In fact, tech support at INFORMIX is looking into what the potential harm might be if we can't solve our problem correctly and have to resort to using this as a crutch fix. Good luck, David Hexem dhexem@deltanet.com