Re: "tbmode -z doesn't kill" - The Answers for Newbies
Posted in 1996
Topics: Server Administration, Jobs, Consulting & Announcements
[Brent Jackson] | Eric Warner <Eric@northop.demon.co.uk> wrote: | | >So opinions range from 'It Doesn't Work' through 'It's Slow' to 'It | >Does Work'. However, there was enough broad agreement in the posts to | >offer some guidelines to the timid: | | >1) Before killing anything, read and understand the flags in tbstat -u. | | >2) If you then issue a tbmode -z, don't be impatient for a result - | >like I was - some posters reported 'hours' before termination. | | >3) Check for process disappearance from tbmode -z, not | >complete disappearence from ps -ef. There can still be 'bits' left | >lying around after Informix has lost interest in the process. | | >4) Don't ever kill -9 an sqlturbo process. EVER! It seems like a fairly | >reliable way to crash your database. | | An excellant summary posting. I missed the original posting - newsfeed | down again :( - so here is my tuppenny bit : | | With OnLine 4.00 we had difficulty with tbmode -z. We found that kill -1 | and -15 were ignored by the process, and kill -9 could bring the engine | down if it hit at the wrong time. However, we found that a kill -8 was a | reliable and safe way of killing the process. | | Note however, since OnLine 4.10 (through to 5.06) we have had no problems | with tbmode -z, though (as already mentioned) it can sometimes take awhile | (~ 30 minutes) to work. Dave Kosenko <davek@informix.com> posted an excellent article on sqlturbo and signals vs. tbmode -z. It was posted on 4 May 93 16:22:15 GMT with Message-ID: <1993May4.162215.21136@informix.com>. I've using the advices in this article since. I don't think they are valid for 6.X and up, though, since there are no pipes in that case. Maybe he can be persuaded to post it again. Also there are some good advice in Joe Lumbley's book "DBA survival guide". Some neat tricks with SIGSTOP! Lastly, I presume the kill -8 is referring to SIGFPE (Floating Point Exeption) as this is signal 8 an all the systems I can lay my hands on. Generally, it is much wiser to use the symbolic names, since some of the signal numbers are different on different platforms. -- 'yvind Gjerstad Systems dept Tollpost-Globe AS N-6301 'ndalsnes/Norway E-mail: ogj@it.tollpost.no Phone: +47 7122 6663 Fax: +47 7122 6694