Killing a cube maintenance job

Question

Hi,
We have a query on the handling of cube maintenance jobs:
If, for some issues with the machine, we kill the AWM application process using the Windows Task Manager Utility while its running a cube build, then the connection to the AW is still active and the Analytic Workspace still remains connected in the Read-Write mode. So the impact is that if we were to restart the cube maintenance job again by re-attaching the AW, after having killed it, then we need to in fact restart the DB instance so that the hanging connection to the AW is done away with.
Is there any way to "completely" kill this connection without having to restart the DB instance?
(The AWM version is use is 10.2.0.3.0A)
Thanks,
Piyush

Answer 1

Something like this will work. You need to run it with sufficient privilege to access the v$session and v$aw_olap tables (sys?) as well as the alter system kill command.
CREATE OR REPLACE PROCEDURE sess_killer(awowner IN VARCHAR2,   awname IN VARCHAR2) AS
mysid NUMBER;
serial NUMBER;
CURSOR c1 IS
SELECT sid,
  serial#
FROM v$session s,
  v$aw_olap a,
  dba_aws d
WHERE d.owner = awowner
AND d.aw_name = awname
AND d.aw_number = a.aw_number
AND a.session_id = s.sid;
sql_stmt VARCHAR2(200);
session_marked
EXCEPTION;
pragma exception_init(session_marked,   -31);
BEGIN
  OPEN c1;
  LOOP
    FETCH c1
    INTO mysid,
      serial;
    EXIT
  WHEN c1 % NOTFOUND;
  --DBMS_OUTPUT.PUT_LINE('Session: ' || mysid || ', ' || serial);
  sql_stmt := 'ALTER system kill SESSION ''' || mysid || ', ' || serial || '''';
-- DBMS_OUTPUT.PUT_LINE(sql_stmt);
  BEGIN
    EXECUTE IMMEDIATE sql_stmt;
  EXCEPTION
  WHEN session_marked THEN
    NULL;
  END;
END LOOP;
CLOSE c1;
END sess_killer;