well, that is understandable. But the question is why it is taking too long to roll back. It is more than 10 hrs now. What is it rolling back for this long? regards - chirag
On 9/15/05, Chen, Sarah <Sarah_Chen@(protected)> wrote: > > I did some research according to this issue before. Here is from my > notes(I > don't know where I got it, but must from one of the Oracle experts): > > when a session is killed, oracle needs to rollback whatever work it had > been > doing - that's why it waits. > > going to the o/s and killing the processes won't speed this up. However, > when you use "alter system kill session", the session doesn't know it's > killed until it makes another request to the database - if it was waiting > for a response (particularly to a remote operation), it may not make a > request for a long time, in which case killing the process from the o/s > may > help to stop the session. at that time, the database still needs to > rollback > whatever had happened already. > > > Sarah > > > -- --Original Message-- -- > From: Yongping Yao [mailto:yaoyongping@(protected)] > Sent: Thursday, September 15, 2005 11:06 AM > To: Chen, Sarah > Cc: Thomas.Mercadante@(protected); chiragdba@(protected); Oracle-L > Freelists > Subject: Re: Session Killed...!! > > > I agree that it's very strange to kill a session in Oracle. As I > experienced, SELECTs are more easier to kill than DMLs and SELECTs > running after a long time are more difficult to kill. UPDATE is the > most difficult to kill. Oracle always reponds as "we have marked the > session killed" but the session is still there and the lock remained > hold. Anyone with good recommendations? > > 2005/9/15, Chen, Sarah <Sarah_Chen@(protected)>: > > > > I always check process id as well as sid and serial# while killing a > session > > from Oracle. > > > > I will always query v$session before and after, and if the status of > > v$session marked "killed", and then I will go ahead kill OS process to > clean > > this session. It always works on Solaris. > > > > Sarah > > > > > > -- --Original Message-- -- > > From: oracle-l-bounce@(protected) > [mailto:oracle-l-bounce@(protected)]On > > Behalf Of Mercadante, Thomas F (LABOR) > > Sent: Thursday, September 15, 2005 10:47 AM > > To: chiragdba@(protected); Oracle-L Freelists > > Subject: RE: Session Killed...!! > > > > > > > > > > If you kill the unix process associated with this session first, and > then > > the session it will go away. > > > > > > > > Killing sessions in Oracle has always been weird. Sometimes they go away > > and sometimes not, depending on OS. In Windows, use orakill to kill the > > session �C this always seems to work in windows. And killing the unix > > session then the Oracle session seems to work just fine in AIX. > > > > > > > > __ ____ ____ ____ ____ ____ ____ > > > > > > From: oracle-l-bounce@(protected) [mailto: > oracle-l-bounce@(protected)] > > On Behalf Of Chirag DBA > > Sent: Thursday, September 15, 2005 10:36 AM > > To: askdba@(protected); Oracle-L Freelists > > Subject: Session Killed...!! > > > > > > > > > > Hi , > > > > > > > > > > > > I saw many users on my database connected for more than 6 days without > any > > activity. > > > > > > > > > > > > I killed 1 user and still the status in v$session is showing as killed. > > > > > > > > > > > > but it it not getting removed from the v$session. > > > > > > > > > > > > I am running 9.0.1.3 <http://9.0.1.3> on solaris. > > > > > > > > > > > > any idea? We already hit the bug ora-07442 (See ora-07442.ora-code.com). > > > > > > > > > > > > regards - chirag > > > -- > Yao Yongping > Learning Oracle, UNIX/Linux... > Love Reading, Classical Music, Philosophy, Economics etc. > Blog: http://blog.csdn.net/ern > http://spaces.msn.com/members/yaoyp/ >
<div>well, that is understandable. But the question is why it is taking too long to roll back.</div> <div> </div> <div>It is more than 10 hrs now.</div> <div> </div> <div>What is it rolling back for this long? </div> <div> </div> <div>regards - chirag<br><br> </div> <div><span class="gmail_quote">On 9/15/05, <b class="gmail_sendername">Chen, Sarah</b> <<a href="mailto:Sarah_Chen@(protected)">Sarah_Chen@(protected)</a>> ; wrote:</span> <blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0 .8ex; BORDER-LEFT: #ccc 1px solid">I did some research according to this issue before. Here is from my notes(I<br>don't know where I got it, but must from one of the Oracle experts): <br><br>when a session is killed, oracle needs to rollback whatever work it had been<br>doing - that's why it waits.<br><br>going to the o/s and killing the processes won't speed this up. However,<br>when you use "alter system kill session", the session doesn't know it's <br>killed until it makes another request to the database - if it was waiting <br>for a response (particularly to a remote operation), it may not make a<br >request for a long time, in which case killing the process from the o/s may <br>help to stop the session. at that time, the database still needs to rollback<br>whatever had happened already.<br><br><br>Sarah<br><br><br>---- -Original Message-- --<br>From: Yongping Yao [mailto:<a href="mailto:yaoyongping @(protected)"> yaoyongping@(protected)</a>]<br>Sent: Thursday, September 15, 2005 11:06 AM<br>To : Chen, Sarah<br>Cc: <a href="mailto:Thomas.Mercadante@(protected)">Thomas .Mercadante@(protected)</a>; <a href="mailto:chiragdba@(protected)"> chiragdba@(protected)</a>; Oracle-L<br>Freelists<br>Subject: Re: Session Killed.. .!!<br><br><br>I agree that it's very strange to kill a session in Oracle. As I <br>experienced, SELECTs are more easier to kill than DMLs and SELECTs <br>running after a long time are more difficult to kill. UPDATE is the<br>most difficult to kill. Oracle always reponds as "we have marked the<br>session killed" but the session is still there and the lock remained <br>hold. Anyone with good recommendations?<br><br>2005/9/15, Chen, Sarah < <a href="mailto:Sarah_Chen@(protected)">Sarah_Chen@(protected)</a>>:<br>><br> > I always check process id as well as sid and serial# while killing a <br>session<br>> from Oracle.<br>><br>> I will always query v$session before and after, and if the status of<br>> v$session marked "killed ", and then I will go ahead kill OS process to<br>clean<br>> this session. It always works on Solaris. <br>><br>> Sarah<br>><br>><br>> -- --Original Message-- --<br> > From: <a href="mailto:oracle-l-bounce@(protected)">oracle-l-bounce @(protected)</a><br>[mailto:<a href="mailto:oracle-l-bounce@(protected)"> oracle-l-bounce@(protected)</a>]On<br>> Behalf Of Mercadante, Thomas F (LABOR)<br>> Sent: Thursday, September 15, 2005 10:47 AM<br>> To: <a href= "mailto:chiragdba@(protected)">chiragdba@(protected)</a>; Oracle-L Freelists <br>> Subject: RE: Session Killed...!!<br>><br>><br>><br>><br> > If you kill the unix process associated with this session first, and then <br>> the session it will go away.<br>><br>><br>><br>> Killing sessions in Oracle has always been weird. Sometimes they go away <br>> and sometimes not, depending on OS. In Windows, use orakill to kill the<br>> session �C this always seems to work in windows.   ;And killing the unix<br>> session then the Oracle session seems to work just fine in AIX. <br>><br>><br>><br>> __ ____ ____ ____ ____ ____ ____<br >><br>><br>> From: <a href="mailto:oracle-l-bounce@(protected)" >oracle-l-bounce@(protected)</a> [mailto:<a href="mailto:oracle-l-bounce @(protected)"> oracle-l-bounce@(protected)</a>]<br>> On Behalf Of Chirag DBA<br>> Sent : Thursday, September 15, 2005 10:36 AM<br>> To: <a href="mailto:askdba @(protected)">askdba@(protected)</a>; Oracle-L Freelists<br>> Subject: Session Killed...!! <br>><br>><br>><br>><br>> Hi ,<br>><br>><br>><br>> <br>><br>> I saw many users on my database connected for more than 6 days without any<br>> activity.<br>><br>><br>><br>><br> ><br>> I killed 1 user and still the status in v$session is showing as killed.<br>><br>><br>><br>><br>><br>> but it it not getting removed from the v$session.<br>><br>><br>><br>><br>> <br>> I am running <a href="http://9.0.1.3">9.0.1.3</a> on solaris.<br>> <br>><br>><br>><br>><br>> any idea? We already hit the bug ora -07442.<br>><br>><br>><br>><br>><br>> regards - chirag <br><br><br>--<br>Yao Yongping<br>Learning Oracle, UNIX/Linux...<br>Love Reading, Classical Music, Philosophy, Economics etc.<br>Blog: <a href="http:/ /blog.csdn.net/ern">http://blog.csdn.net/ern</a><br> <a href="http://spaces.msn.com/members/yaoyp/"> http://spaces.msn.com/members/yaoyp/</a><br></blockquote></div><br>
