Reputation
980
Top tag
Next privilege 1,000 Rep.
See votes, expandable usercard
Badges
3 10
Newest
 Yearling
Impact
~78k people reached

  • 0 posts edited
  • 1 helpful flag
  • 45 votes cast
Apr
26
comment Oracle session killed, but the table still not released
long story short, what was done must be undone. check v$session, it's transaction. in v$transaction you will find number of undo blocks. When it gets to zero, the killed session will disappear.
Apr
8
revised Is there an Oracle function to check for valid Japanese characters (Non-ASCII)?
added 342 characters in body
Apr
8
answered Is there an Oracle function to check for valid Japanese characters (Non-ASCII)?
Mar
31
answered Updating data from subpartition when the subpartition name is unknown
Mar
31
comment Why is installing ORACLE ODBC client a complex task?
But on the other side Oracle forbids re-distribution of client drivers. Unless you're Oracle business partner.
Mar
2
comment SCAN listener redirects connection to another port
Check which listener name is used dispatchers parameter DISPATCHERS="(PROTOCOL=tcp)(LISTENER=listener_alias)" If this does not work for you I'm afraid the best option is to open a SR with Oracle.
Feb
26
revised ORA-04031: unable to allocate 32 bytes of shared memory
added 1076 characters in body
Feb
3
comment Oracle over-indexed tables
Too many indexes can cause problems with exec plan stability. .i.e. you may face problems, when Oracle suddenly "change exec plan without warning". It can also be sign, that there were performance problems in the past and somebody tried to fix them, without knowing what exactly is going on.
Feb
1
awarded  Yearling
Jan
6
comment Oracle, Cursor_Sharing = Force and Queries with literal comparisons forcing FTS
sharing was supposed to improve the performance. But now you face the negative impact of this "improvement". The best thing you can do is to contact the software vendor.
Jan
6
comment Oracle, Cursor_Sharing = Force and Queries with literal comparisons forcing FTS
Theoretically the predicate 'a' = 'b' (or 'a' = 'a' ) could be elimanated as tautology at early stage of the parsing. But it looks like the bind variable replacement (cursor sharing) takes place before. You can not do too much with that. Google the terms "bind variable peeking" and "SQL plan baseline". The most important rule for you is that a single sql_id (a sql statement hash) can have only one execution plan. The whole idea with the 'or' predicate is sick. You WANT full table scan if this predicate evaluates to true, while you want index access when it evalutates to false. IMHO the cursor
Nov
12
comment Unable to turn off automatic memory management (AMM)
You stored our changes into spifile, but you use initdev12.ora to start the database. What is in this file?
Nov
2
comment adrci not purge trace files and alert logs
AFAIK alert log and listener.log can not be managed by adrci
Oct
27
comment Select performance problems with conditional index vs non-conditional on large volume table
Try to execute the query with gather_stats hint and then use dbms_xplan to show execution plan. This might give you some numbers describing how your query is evaluated in reality.
Oct
27
awarded  Custodian
Oct
27
reviewed Approve Is Overwriting the PFILE with the database on dangerous?
Oct
26
comment Select performance problems with conditional index vs non-conditional on large volume table
What is the exec plan? Index range scan or fast full scan? Do you have a license for real time SQL monitoring? Did you try to execute the query with gather_stats hint?
Oct
26
answered Is Overwriting the PFILE with the database on dangerous?
Oct
21
comment How to optimize very slow SELECT with LEFT JOINs over big tables
OT: isn't it strange that you use LIMIT withour ORDER BY? This will return some random 100000 rows?
Oct
9
comment How to change dbid in oracle to specific value?
AFAIK the order should be execute rman; set dbid; connect target / catalog ...; restore controlfile; startup mount;. The DBID must not be read from controlfile.