Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Oracle screws up Sun host

Re: Oracle screws up Sun host

From: Anton Buijs <remove_aammbuijs_at_xs4all.nl>
Date: Mon, 21 Apr 2003 21:48:40 +0200
Message-ID: <3ea44b13$0$49113$e4fe514c@news.xs4all.nl>

Rick Denoire <100.17706_at_germanynet.de> schreef in berichtnieuws 9ef8avkappqgv0ub05rt2bvia2u2gd7c5e_at_4ax.com...
| We are running two Oracle DB (both 8.1.7) on the same host (Sun
| E3500/Solaris 7). The first is small and stable and is never changed
| in its structure (it is a commercial product) - it should run 24x7.
| The second one is "big" (60 GB Data + Indexes), is being changed
| almost constantly and runs PL/SQL jobs lasting many hours (Data Mining
| operations). This second DB produces about 3 ORA-00600 errors a week
| and at times, it MUST be shut down (because of runaway processes, for
| example). Normally, shutting down the "big" DB (the instance) is
| tolerable.
|
| But if it must be shut down due to a problem, if often can't be. Using
| "shutdown immediate" goes as far as closing the DB, but then, the
| process of shutting down the DB hangs. If after a while I interrupt
| the process with Control-C, I get the SQL prompt but I am out of
| business (disconnected). Nothing more can be done with the DB, since I
| can't even connect ("instance is down"). Then I notice that there are
| still one or more Oracle background processes running, which I have to
| kill at the OS command line. Then I am still unable to connect. As a
| last effort, I remove the file $ORACLE_HOME/dbs/lk<SID-NAME>, which
| does not help either. Somehow the host is crewed up at this point, and
| the only chance I have to get things fixed is to restart the whole
| host.
|
| This is a shame since this is not windows and there should be a way of
| cleaning up things without having to restart the host. Restarting the
| host is not tolerable (it runs the 1st 24x7 DB and an Intranet Server,
| and it is a NIS Master).
|
| So your advice will be to separate the DBs. We are actually planing
| exactly that, the new server hardware is being ordered. But my
| question still remains, how to let the other instances untouched when
| one of them goes nuts. >>The new server will hold several instances<<
| Hm.
|
| Any clue?
|
| Rick Denoire

Try to use shutdown abort. I don't see a reason for removing the lk<sid> file.
Do kill -15 to kill any remaining processes of the big database, if it does not help use kill -9 as a last resort.
Why do you conclude "the host is screwed up" and needs a reboot?

Further I would certainly look for the trace files produced by the ORA-600 errors. There is a utility on Metalink where you can load the dump stack and get possible known problems reported. If you are lucky you find an existing bug fix is available. When you don't find a solution, open a TAR with Oracle support. ORA-600 errors can only be resolved with a TAR. Received on Mon Apr 21 2003 - 14:48:40 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US