Comments on: Alter system kill session and ORA-00031: session marked for kill

By: Tanel Poder

Tanel Poder — Wed, 10 Mar 2010 00:08:47 +0000

@ronald Ronald, you'll probably find that the session got finally "killed" after 100 seconds, the time you requested it to sleep. But if the sleep is longer, or the session is hopelessly stuck in some loop (due a bug) then the session won't be killed until the sleep/operation completes.

By: ronald

ronald — Tue, 09 Mar 2010 10:07:29 +0000

It worked for me!
to test if you can kill a session

1. create the session by logging into the database as user x and execute exec dbms_lock.sleep(100)
2. create another session and select sid,serial#,event from v$session where username=’userx’;
3. on the same session, ‘alter system kill session ‘sid,serial#’;
BIG NOTE: Wait for 60 seconds and don’t be impatient. It’ll be killed after sometime.

By: Killing Database Sessions – tips from Tanel on killing at the OS level & using oradebug « Kirk Brocas' Oracle Stuff

Wed, 03 Mar 2010 21:33:28 +0000

[...] Alter system kill session and ORA-00031: session marked for kill Categories: DBA, oracle Tags: oracle, oradebug, session management, sessions Comments (0) Trackbacks (0) Leave a comment Trackback [...]

By: Online Car Insurance >> http://onlinecarinsuranceclaims.com/

Online Car Insurance >> http://onlinecarinsuranceclaims.com/ — Mon, 23 Nov 2009 20:40:41 +0000

[... - blog.tanelpoder.com is another nice website of information. Car insurance claims [... -

By: Raj

Raj — Wed, 07 Oct 2009 19:43:30 +0000

very good article and informative. Even in 10g kill immediate is available. As you said some times it works .. some times it doesn’t.

By: Tanel Poder

Tanel Poder — Wed, 26 Aug 2009 06:12:05 +0000

Stay tuned for another article on this topic!

By: Tanel Poder

Tanel Poder — Wed, 26 Aug 2009 06:11:39 +0000

I will write a separate article about the disconnect session and kill session immediate. In my previous experience the immediate options haven’t worked reliably. I tested on 11.1.0.7 on Linux and still it succeeds only in some cases.

Regarding why I would not want to kill -9 a session when its already rolling back a long transaction – the reason is that I would prefer the session to clean up after itself, instead of disturbing it and letting smon to take over.

There can be value in killing the session and letting smon to take over though, if you’re on EE, have lots of CPUs and have fast_start_parallel_rollback enabled, this can make smon spawn multiple slaves for recovery. However it doesn’t always necessarily speed up the rollback and I’ve had bad experiences with completely hung parallel rollbacks in 9.2. I’m sure this issue is fixed by now but I’m just being careful – if a session is already rolling back after itself and if the estimated rollback time is acceptable, then don’t touch it!

By: Thyge G. Jacobsen

Thyge G. Jacobsen — Tue, 18 Aug 2009 11:13:02 +0000

… and 9i http://download.oracle.com/docs/cd/B10501_01/server.920/a96540/statements_23a.htm#SQLRF00902.

By: Entradas de Oracle semana 33 « Gruñidos sobre Oracle y SAP

Entradas de Oracle semana 33 « Gruñidos sobre Oracle y SAP — Sun, 16 Aug 2009 18:43:40 +0000

[...] 7.- Tanel Poder visita los entresijos del ORA-00031: session marked for kill. [...]

By: Dave Herring

Dave Herring — Fri, 14 Aug 2009 20:34:37 +0000

Great info as always, Tanel!

Correct me if I’m wrong, but under 9i it seems that after you killed an Oracle session and it gets “marked for kill”, the process goes away but not the session. So at that point you can’t get the SPID, as it’s not found in V$PROCESS.

I believe this ISN’T the case in 10g. I need to fit 9i on my laptop and test …