How to block reading in "read commited"?

I wrote following code in my sp:
select id into tempflag from table where conditions....
if tempflag is null then
insert xx into table
else
do something else
end if;
commit;
Now let's assume this sp is excuted 2 times at almost same time, 1st transaction has inserted data but not commited yet, and 2nd transaction still not able to read the new inserted data, so it inserted data again. The result is I got 2 data in table, which is what I don't want to see.
What I want is, when 1st transaction not commited yet, 2nd transaction should be blocking at the select statement, this should prevent the second time inserting.
Actually this is the default behavior in MSSqlserver(read commited), but I found it quite different in Oracle.
So how can I implement the above idea? Or any other better solutions? I'm new to Oracle.

874717 wrote:
I wrote following code in my sp:
select id into tempflag from table where conditions....
if tempflag is null then
insert xx into table
else
do something else
end if;
commit;
Now let's assume this sp is excuted 2 times at almost same time, 1st transaction has inserted data but not commited yet, and 2nd transaction still not able to read the new inserted data, so it inserted data again. The result is I got 2 data in table, which is what I don't want to see.
What I want is, when 1st transaction not commited yet, 2nd transaction should be blocking at the select statement, this should prevent the second time inserting.
Actually this is the default behavior in MSSqlserver(read commited), but I found it quite different in Oracle.
So how can I implement the above idea? Or any other better solutions? I'm new to Oracle.Just put a unique constraint in ID column of your table. I guess that will do the job for you.
Following piece of code is not necessory
select id into tempflag from table where conditions....
if tempflag is null thenDont try to select the table and check if the ID already exist. Just have a unique constraint defined and perform your insert.

Similar Messages

How to block others from reading email?

How to block others from reading email on the Ipad 2?

iPads are not multi user devices, they are either single user devices with privacy by setting the mentioned passcode to protect entry into the device or it is a shared device where everything on the iPad is available to everyone with access to the iPad.

How to block order for a particular customer. pls read clearly

hi
some customers may be asked for cash and carry procedure.
bcos they do very small sales they have access to only cash sale document ie CS.
here how to block OR, RO etc to that particular customer or group of customers
regards
sridhar

Hi,
One simple way would be to give those customers a special sales area (e.g. a separate distribution channel) than for normal customers. Then you do not allow the Document Type for this sales area in the transaction "Assign Sales Areas to Sales Document Types" in IMG (Sales & Distribution -> Sales -> Sales Documents -> Sales Documents Header).
Ensure that the customer is created to this "blocked" sales area only. Then when you try to create the sales order, system would give an error message saying customer is not defined for the sales area, and would not let you proceed.
Hope this helps,
Regards
Nikhilesh

Read committed isolation level must not produce nonrepeatable read

Hi,
I am sql server DBA. But i am trying to improve myself in oracle too.
I read isolation levels in oracle. And it says, In read committed isolation level , oracle guarantees the result set that contains committed records at the beginning of reading operation.
İf it is guaranteed , how does nonrepeatable read can occur? It must not occur then.
I think , I misunderstood something .
can you explain me?
Thanks

>
I read isolation levels in oracle. And it says, In read committed isolation level , oracle guarantees the result set that contains committed records at the beginning of reading operation.
İf it is guaranteed , how does nonrepeatable read can occur? It must not occur then.
>
See the 'Multiversion Concurrency Control' section in the database concepts doc. It discusses this and has a simple diagram (can't post it) that shows it.
http://docs.oracle.com/cd/B28359_01/server.111/b28318/consist.htm
>
As a query enters the execution stage, the current system change number (SCN) is determined. In Figure 13-1, this system change number is 10023. As data blocks are read on behalf of the query, only blocks written with the observed SCN are used. Blocks with changed data (more recent SCNs) are reconstructed from data in the rollback segments, and the reconstructed data is returned for the query. Therefore, each query returns all committed data with respect to the SCN recorded at the time that query execution began. Changes of other transactions that occur during a query's execution are not observed, guaranteeing that consistent data is returned for each query.
Statement-Level Read Consistency
Oracle Database always enforces statement-level read consistency. This guarantees that all the data returned by a single query comes from a single point in time—the time that the query began. Therefore, a query never sees dirty data or any of the changes made by transactions that commit during query execution. As query execution proceeds, only data committed before the query began is visible to the query. The query does not see changes committed after statement execution begins.
>
The first sentence is the key:
>
As a query enters the execution stage, the current system change number (SCN) is determined.
>
Oracle will only query data AS OF that SCN that was determined.
If you now rerun the query Oracle repeats the process: it determines the SCN again which could be newer if other users have committed changes.
That second execution of the query may find that some rows have been modified or even deleted and that new rows have been inserted: nonrepeatable read.
If you use the SERIALIZABLE isolation level then that second query will use the SCN that was determined at the very START of the transaction. For the simple example above it means the second query would use the SAME SCN that the first query used: so the same data would be returned.
Table 13-2 in that doc (a few pages down) lists the isolation levels
>
Read committed
This is the default transaction isolation level. Each query executed by a transaction sees only data that was committed before the query (not the transaction) began. An Oracle Database query never reads dirty (uncommitted) data.
Because Oracle Database does not prevent other transactions from modifying the data read by a query, that data can be changed by other transactions between two executions of the query. Thus, a transaction that runs a given query twice can experience both nonrepeatable read and phantoms.

Single-statement 'write consistency' on read committed?

Please note that in the following I'm only concerned about single-statement read committed transactions. I do realize that for a multi-statement read committed transaction Oracle does not guarantee transaction set consistency without techniques like select for update or explicit hand-coded locking.
According to the documentation Oracle guarantees 'statement-level transaction set consistency' for queries in read committed transactions. In many cases, Oracle also provides single-statement write consistency. However, when an update based on a consistent read tries to overwrite changes committed by other transactions after the statement started, it creates a write conflict. Oracle never reports write conflicts on read committed. Instead, it automatically handles them based on the new values for the target table columns referenced by the update.
Let's consider a simple example. Again, I do realize that the following design might look strange or even sloppy, but the ability to produce a quality design when needed is not an issue here. I'm simply trying to understand the Oracle's behavior on write conflicts in a single-statement read committed transaction.
A valid business case behind the example is rather common - a financial institution with two-stage funds transfer processing. First, you submit a transfer (put transfer amounts in the 'pending' column of the account) in case the whole financial transaction is in doubt. Second, after you got all the necessary confirmations you clear all the pending transfers making the corresponding account balance changes, resetting pending amount and marking the accounts cleared by setting the cleared date. Neither stage should leave the data in inconsistent state: sum (amount) for all rows should not change and the sum (pending) for all rows should always be 0 on either stage:
Setup:
create table accounts
acc int primary key,
amount int,
pending int,
cleared date
Initially the table contains the following:
ACC AMOUNT PENDING CLEARED
1 10 -2
2 0 2
3 0 0 26-NOV-03
So, there is a committed database state with a pending funds transfer of 2 dollars from acc 1 to acc 2. Let's submit another transfer of 1 dollar from acc 1 to acc 3 but do not commit it yet in SQL*Plus Session 1:
update accounts
set pending = pending - 1, cleared = null where acc = 1;
update accounts
set pending = pending + 1, cleared = null where acc = 3;
ACC AMOUNT PENDING CLEARED
1 10 -3
2 0 2
3 0 1
And now let's clear all the pending transfers in SQL*Plus Session 2 in a single-statement read-committed transaction:
update accounts
set amount = amount + pending, pending = 0, cleared = sysdate
where cleared is null;
Session 2 naturally blocks. Now commit the transaction in session 1. Session 2 readily unblocks:
ACC AMOUNT PENDING CLEARED
1 7 0 26-NOV-03
2 2 0 26-NOV-03
3 0 1
Here we go - the results produced by a single-statement transaction read committed transaction in session 2, are inconsistent � the second funds transfer has not completed in full. Session 2 should have produced the following instead:
ACC AMOUNT PENDING CLEARED
1 7 0 26-NOV-03
2 2 0 26-NOV-03
3 1 0 26-NOV-03
Please note that we would have gotten the correct results if we ran the transactions in session 1 and session 2 serially. Please also note that no update has been lost. The type of isolation anomaly observed is usually referred to as a 'read skew', which is a variation of 'fuzzy read' a.k.a. 'non-repeatable read'.
But if in the session 2 instead of:
-- scenario 1
update accounts
set amount = amount + pending, pending = 0, cleared = sysdate
where cleared is null;
we issued:
-- scenario 2
update accounts
set amount = amount + pending, pending = 0, cleared = sysdate
where cleared is null and pending <> 0;
or even:
-- scenario 3
update accounts
set amount = amount + pending, pending = 0, cleared = sysdate
where cleared is null and (pending * 0) = 0;
We'd have gotten what we really wanted.
I'm very well aware of the 'select for update' or serializable il solution for the problem. Also, I could present a working example for precisely the above scenario for a major database product, providing the results that I would consider to be correct. That is, the interleaving execution of the transactions has the same effect as if they completed serially. Naturally, no extra hand-coded locking techniques like select for update or explicit locking is involved.
And now let's try to understand what just has happened. Playing around with similar trivial scenarios one could easily figure out that Oracle clearly employs different strategies when handling update conflicts based on the new values for the target table columns, referenced by the update. I have observed the following cases:
A. The column values have not changed: Oracle simply resumes using the current version of the row. It's perfectly fine because the database view presented to the statement (and hence the final state of the database after the update) is no different from what would have been presented if there had been no conflict at all.
B. The row (including the columns being updated) has changed, but the predicate columns haven't (see scenario 1): Oracle resumes using the current version of the row. Formally, this is acceptable too as the ANSI read committed by definition is prone to certain anomalies anyway (including the instance of a 'read skew' we've just observed) and leaving behind somewhat inconsistent data can be tolerated as long as the isolation level permits it. But please note - this is not a 'single-statement write consistent' behavior.
C. Predicate columns have changed (see scenario 2 or 3): Oracle rolls back and then restarts the statement making it look as if it did present a consistent view of the database to the update statement indeed. However, what seems confusing is that sometimes Oracle restarts when it isn't necessary, e.g. when new values for predicate columns don't change the predicate itself (scenario 3). In fact, it's bit more complicated � I also observed restarts on some index column changes, triggers and constraints change things a bit too � but for the sake of simplicity let's no go there yet.
And here come the questions, assuming that (B) is not a bug, but the expected behavior:
1. Does anybody know why it's never been documented in detail when exactly Oracle restarts automatically on write conflicts once there are cases when it should restart but it won't? Many developers would hesitate to depend on the feature as long as it's not 'official'. Hence, the lack of the information makes it virtually useless for critical database applications and a careful app developer would be forced to use either serializable isolation level or hand-coded locking for a single-statement update transaction.
If, on the other hand, it's been documented, could anybody please point me to the bit in the documentation that:
a) Clearly states that Oracle might restart an update statement in a read committed transaction because otherwise it would produce inconsistent results.
b) Unambiguously explains the circumstances when Oracle does restart.
c) Gives clear and unambiguous guidelines on when Oracle doesn't restart and therefore when to use techniques like select for update or the serializable isolation level in a single-statement read committed transaction.
2. Does anybody have a clue what was the motivation for this peculiar design choice of restarting for a certain subset of write conflicts only? What was so special about them? Since (B) is acceptable for read committed, then why Oracle bothers with automatic restarts in (C) at all?
3. If, on the other hand, Oracle envisions the statement-level write consistency as an important advantage over other mainstream DBMSs as it clear from the handling of (C), does anybody have any idea why Oracle wouldn't fix (B) using well-known techniques and always produce consistent results?

I intrigued that this posting has attracted so little interest. The behaviour described is not intuitive and seems to be undocumented in Oracle's manuals.
Does the lack of response indicate:
(1) Nobody thinks this is important
(2) Everybody (except me) already knew this
(3) Nobody understands the posting
For the record, I think it is interesting. Having spent some time investigating this, I believe the described is correct, consistent and understandable. But I would be happier if Oracle documented in the Transaction sections of the Manual.
Cheers, APC

( if snapshot isolation level is enable in db.) Is the version chain is generated if any read commited transation is executed or it only gnerates version chain, if any snapshot transaction is in running status.

hi,
I have enable snapshot isolation level, in my database all queries execute in read commited isolation only one big transaction uses snapshot isolation.
q1)i wanted to know if snapshot silation transaction is not running but database is enable for snapshot ,then will the normal
queries using read commited will create versions or not.
yours sincerley.

Enabling snapshot isolation level on DB level does not change behavior of queries in any other isolation levels. In that option you are eliminating all blocking even between writers (assuming they do not update
the same rows) although it could lead to 3960 errors (data has been modified by other sessions).
Best Regards,Uri Dimant SQL Server MVP,
http://sqlblog.com/blogs/uri_dimant/
MS SQL optimization: MS SQL Development and Optimization
MS SQL Consulting:
Large scale of database and data cleansing
Remote DBA Services:
Improves MS SQL Database Performance
SQL Server Integration Services:
Business Intelligence

Can't update while select is active in read-committed mode

Hi
This is my business logic:
I m selecting the values from three tables by UNION ALL and i m geting the record set in while loop at the same time i try to insert the values which is getting from select statement in the First Table in the same while loop.
my problem is when i try to insert into the same table the following error is occured
"can't update while select is active in read-committed mode"
Now what can i do ? please give me ur solution ...
Thankx
Merlina

depending on how many values you have, you could store the results in memory, a vector or similar collection.
Then close the select statement, and insert the values from your vector into the table.... should work then.
Or, there may be a way of creating an updatable resultset ?
This question is better directed at the
"Java Database Connectivity (JDBC) & Transactions (JTA/JTS)" section of the forum. Maybe post a questions there, ensuring you give the URL of this question here too to avoid cross-posting.
regards,
Owen

Unable to relate consistent reads with number of phsyical block reads

HI ,
The question is we have observed the consistent reads are much more than total buffers required to give the results back.
I have flushed the buffer_cache before executing the query and also queried the V$BH to know the buffer details for these objects ...after the flush before firing the query we don't have any buffers regarding these tables. Which is expected.
We are doing DB file sequential reads through the plan and it will result into a single block read at a time.
Please take a close look at "TABLE ACCESS BY INDEX ROWID CMPGN_DIM (cr=45379 pr=22949 pw=0 time=52434931 us)" line in the below row source plan..
Here we have only 22949 physical reads means 22949 data buffers but we are seeing 45379 consistent gets.
Note: We have the CMPGN_DIM and AD_GRP tables are in 4M block size table space and we have only than the default db_cache_size . My database block size is 8192.
Can you please help me in understand how the 22949 sequential reads result into 45379 consistant gets.
Even the V$BH query buffer details matches with physical reads .
query row source plan from 10043 trace:
27 SORT ORDER BY (cr=92355 pr=47396 pw=0 time=359030364 us)
27 WINDOW SORT (cr=92355 pr=47396 pw=0 time=359030088 us)
27 NESTED LOOPS OUTER (cr=92355 pr=47396 pw=0 time=359094569 us)
27 NESTED LOOPS OUTER (cr=92276 pr=47395 pw=0 time=359041825 us)
27 VIEW (cr=92197 pr=47393 pw=0 time=358984314 us)
27 UNION-ALL (cr=92197 pr=47393 pw=0 time=358984120 us)
26 HASH GROUP BY (cr=92197 pr=47393 pw=0 time=358983665 us)
9400 VIEW (cr=92197 pr=47393 pw=0 time=359094286 us)
9400 COUNT (cr=92197 pr=47393 pw=0 time=359056676 us)
9400 VIEW (cr=92197 pr=47393 pw=0 time=359009672 us)
9400 SORT ORDER BY (cr=92197 pr=47393 pw=0 time=358972063 us)
9400 HASH JOIN OUTER (cr=92197 pr=47393 pw=0 time=358954170 us)
9400 VIEW (cr=92191 pr=47387 pw=0 time=349796124 us)
9400 HASH JOIN (cr=92191 pr=47387 pw=0 time=349758517 us)
94 TABLE ACCESS BY INDEX ROWID CMPGN_DIM (cr=45379 pr=22949 pw=0 time=52434931 us)
50700 INDEX RANGE SCAN IDX_CMPGN_DIM_UK1 (cr=351 pr=349 pw=0 time=1915239 us)(object id 55617)
60335 TABLE ACCESS BY INDEX ROWID AD_GRP (cr=46812 pr=24438 pw=0 time=208234661 us)
60335 INDEX RANGE SCAN IDX_AD_GRP2 (cr=613 pr=611 pw=0 time=13350221 us)(object id 10072801)
7 VIEW (cr=6 pr=6 pw=0 time=72933 us)
7 HASH GROUP BY (cr=6 pr=6 pw=0 time=72898 us)
162 PARTITION RANGE SINGLE PARTITION: 4 4 (cr=6 pr=6 pw=0 time=45363 us)
162 PARTITION HASH SINGLE PARTITION: 676 676 (cr=6 pr=6 pw=0 time=44690 us)
162 INDEX RANGE SCAN PK_AD_GRP_DTL_FACT PARTITION: 3748 3748 (cr=6 pr=6 pw=0 time=44031 us)(object id 8347241)
1 FAST DUAL (cr=0 pr=0 pw=0 time=9 us)
25 TABLE ACCESS BY INDEX ROWID AD_GRP (cr=79 pr=2 pw=0 time=29817 us)

I think that I understand your question. The consistent gets statistic (CR) indicates the number of times blocks were accessed in memory, and doing so possibly required undo to be applied to the blocks to provide a consistent get. The physical read statistic (PR) indicates the number of blocks that were accessed from disk. The consistent gets statistic may be very close to the physical reads statistic, or very different depending on several factors. A test case might best explain why the CR and PR statistics may differ significantly. First, creating the test objects:
CREATE TABLE T1 AS
SELECT
ROWNUM C1,
1000000-ROWNUM C2,
RPAD(TO_CHAR(ROWNUM),800,'X') C3
FROM
DUAL
CONNECT BY
LEVEL<=1000000;
CREATE INDEX INT_T1_C1 ON T1(C1);
CREATE INDEX INT_T1_C2 ON T1(C2);
CREATE TABLE T2 AS
SELECT
ROWNUM C1,
1000000-ROWNUM C2,
RPAD(TO_CHAR(ROWNUM),800,'X') C3
FROM
DUAL
CONNECT BY
LEVEL<=100000;
COMMIT;
EXEC DBMS_STATS.GATHER_TABLE_STATS(OWNNAME=>USER,TABNAME=>'T1',CASCADE=>TRUE,ESTIMATE_PERCENT=>NULL)
EXEC DBMS_STATS.GATHER_TABLE_STATS(OWNNAME=>USER,TABNAME=>'T2',CASCADE=>TRUE,ESTIMATE_PERCENT=>NULL)We now have 2 tables, the first with 1,000,000 rows with about 8 rows per block and having 2 indexes, and the second table with 100,000 rows and no indexes.
SELECT
TABLE_NAME,
PCT_FREE,
NUM_ROWS,
BLOCKS
FROM
USER_TABLES
WHERE
TABLE_NAME IN ('T1','T2');
TABLE_NAME   PCT_FREE   NUM_ROWS     BLOCKS
T1                 10    1000000     125597
T2                 10     100000      12655
COLUMN INDEX_NAME FORMAT A10
SELECT
INDEX_NAME,
BLEVEL,
LEAF_BLOCKS,
DISTINCT_KEYS DK,
CLUSTERING_FACTOR CF,
NUM_ROWS
FROM
USER_INDEXES
WHERE
TABLE_NAME IN ('T1','T2');
INDEX_NAME     BLEVEL LEAF_BLOCKS         DK         CF   NUM_ROWS
INT_T1_C1           2        2226    1000000     125000    1000000
INT_T1_C2           2        2226    1000000     125000    1000000Now a test script to try a couple experiments with the two tables:
SET LIN 120
SET AUTOTRACE TRACEONLY STATISTICS EXPLAIN
SET TIMING ON
SPOOL C:\MYTEST.TXT
ALTER SESSION SET STATISTICS_LEVEL=TYPICAL;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'TEST1';
ALTER SESSION SET EVENTS '10046 TRACE NAME CONTEXT FOREVER, LEVEL 8';
SELECT /*+ USE_HASH(T1 T2) */
T1.C1,
T2.C2,
T1.C3
FROM
T1,
T2
WHERE
T1.C2=T2.C2
AND T1.C2 BETWEEN 900000 AND 1000000;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'TEST2';
SELECT /*+ USE_NL(T1 T2) */
T1.C1,
T2.C2,
T1.C3
FROM
T1,
T2
WHERE
T1.C2=T2.C2
AND T1.C2 BETWEEN 900000 AND 1000000;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'TEST3';
SELECT /*+ USE_HASH(T1 T2) */
T1.C1,
T2.C2,
T1.C3
FROM
T1,
T2
WHERE
T1.C1=T2.C1
AND T1.C1 BETWEEN 1 AND 100000;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SYSTEM FLUSH BUFFER_CACHE;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'TEST4';
SELECT /*+ USE_NL(T1 T2) */
T1.C1,
T2.C2,
T1.C3
FROM
T1,
T2
WHERE
T1.C1=T2.C1
AND T1.C1 BETWEEN 1 AND 100000;
ALTER SESSION SET TRACEFILE_IDENTIFIER = 'TEST5';
SELECT /*+ USE_NL(T1 T2) FIND_ME */
T1.C1,
T2.C2,
T1.C3
FROM
T1,
T2
WHERE
T1.C1=T2.C1
AND T1.C1 BETWEEN 1 AND 100000;
SET AUTOTRACE OFF
ALTER SESSION SET EVENTS '10046 TRACE NAME CONTEXT OFF';
SPOOL OFFTest script output follows (note that the script was executed twice so that statistics related to the hard parse would be excluded):
SQL> SELECT /*+ USE_HASH(T1 T2) */
2    T1.C1,
3    T2.C2,
4    T1.C3
5 FROM
6    T1,
7    T2
8 WHERE
9    T1.C2=T2.C2
10    AND T1.C2 BETWEEN 900000 AND 1000000;
100000 rows selected.
Elapsed: 00:00:22.65
Execution Plan
Plan hash value: 488978626
| Id | Operation                    | Name      | Rows | Bytes |TempSpc| Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |           | 99999 |    77M|       | 20139   (1)| 00:04:02 |
|* 1 | HASH JOIN                   |           | 99999 |    77M| 1664K| 20139   (1)| 00:04:02 |
|* 2 |   TABLE ACCESS FULL          | T2        |   100K|   488K|       | 3435   (1)| 00:00:42 |
|   3 |   TABLE ACCESS BY INDEX ROWID| T1        |   100K|    77M|       | 12733   (1)| 00:02:33 |
|* 4 |    INDEX RANGE SCAN          | INT_T1_C2 |   100K|       |       |   226   (1)| 00:00:03 |
Predicate Information (identified by operation id):
   1 - access("T1"."C2"="T2"."C2")
   2 - filter("T2"."C2">=900000 AND "T2"."C2"<=1000000)
   4 - access("T1"."C2">=900000 AND "T1"."C2"<=1000000)
Statistics
          0 recursive calls
          0 db block gets
      37721 consistent gets
      25226 physical reads
          0 redo size
   82555058 bytes sent via SQL*Net to client
      73722 bytes received via SQL*Net from client
       6668 SQL*Net roundtrips to/from client
          0 sorts (memory)
          0 sorts (disk)
     100000 rows processed
STAT lines from the 10046 trace:
STAT #37 id=1 cnt=100000 pid=0 pos=1 obj=0 op='HASH JOIN (cr=37721 pr=25226 pw=0 time=106305676 us)'
STAT #37 id=2 cnt=100000 pid=1 pos=1 obj=48144 op='TABLE ACCESS FULL T2 (cr=12511 pr=12501 pw=0 time=13403966 us)'
STAT #37 id=3 cnt=100000 pid=1 pos=2 obj=48141 op='TABLE ACCESS BY INDEX ROWID T1 (cr=25210 pr=12725 pw=0 time=103903740 us)'
STAT #37 id=4 cnt=100000 pid=3 pos=1 obj=48143 op='INDEX RANGE SCAN INT_T1_C2 (cr=6877 pr=225 pw=0 time=503602 us)'Elapsed: 00:00:00.01
SQL> SELECT /*+ USE_NL(T1 T2) */
2    T1.C1,
3    T2.C2,
4    T1.C3
5 FROM
6    T1,
7    T2
8 WHERE
9    T1.C2=T2.C2
10    AND T1.C2 BETWEEN 900000 AND 1000000;
100000 rows selected.
Elapsed: 00:00:20.17
Execution Plan
Plan hash value: 1773329022
| Id | Operation                   | Name      | Rows | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |           | 99999 |    77M|   303K (1)| 01:00:43 |
|   1 | TABLE ACCESS BY INDEX ROWID| T1        |     1 |   810 |     3   (0)| 00:00:01 |
|   2 |   NESTED LOOPS              |           | 99999 |    77M|   303K (1)| 01:00:43 |
|* 3 |    TABLE ACCESS FULL        | T2        |   100K|   488K| 3435   (1)| 00:00:42 |
|* 4 |    INDEX RANGE SCAN         | INT_T1_C2 |     1 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - filter("T2"."C2">=900000 AND "T2"."C2"<=1000000)
   4 - access("T1"."C2"="T2"."C2")
       filter("T1"."C2">=900000 AND "T1"."C2"<=1000000)
Statistics
          0 recursive calls
          0 db block gets
     250219 consistent gets
      25227 physical reads
          0 redo size
   82555058 bytes sent via SQL*Net to client
      73722 bytes received via SQL*Net from client
       6668 SQL*Net roundtrips to/from client
          0 sorts (memory)
          0 sorts (disk)
     100000 rows processed
STAT lines from the 10046 trace:
STAT #36 id=1 cnt=100000 pid=0 pos=1 obj=48141 op='TABLE ACCESS BY INDEX ROWID T1 (cr=250219 pr=25227 pw=0 time=61410637 us)'
STAT #36 id=2 cnt=200001 pid=1 pos=1 obj=0 op='NESTED LOOPS (cr=231886 pr=12727 pw=0 time=3000840 us)'
STAT #36 id=3 cnt=100000 pid=2 pos=1 obj=48144 op='TABLE ACCESS FULL T2 (cr=18344 pr=12501 pw=0 time=14103896 us)'
STAT #36 id=4 cnt=100000 pid=2 pos=2 obj=48143 op='INDEX RANGE SCAN INT_T1_C2 (cr=213542 pr=226 pw=0 time=1929742 us)'
SQL> SELECT /*+ USE_HASH(T1 T2) */
2    T1.C1,
3    T2.C2,
4    T1.C3
5 FROM
6    T1,
7    T2
8 WHERE
9    T1.C1=T2.C1
10    AND T1.C1 BETWEEN 1 AND 100000;
100000 rows selected.
Elapsed: 00:00:20.35
Execution Plan
Plan hash value: 689276421
| Id | Operation                    | Name      | Rows | Bytes |TempSpc| Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT             |           | 99999 |    77M|       | 20143   (1)| 00:04:02 |
|* 1 | HASH JOIN                   |           | 99999 |    77M| 2152K| 20143   (1)| 00:04:02 |
|* 2 |   TABLE ACCESS FULL          | T2        |   100K|   976K|       | 3435   (1)| 00:00:42 |
|   3 |   TABLE ACCESS BY INDEX ROWID| T1        |   100K|    76M|       | 12733   (1)| 00:02:33 |
|* 4 |    INDEX RANGE SCAN          | INT_T1_C1 |   100K|       |       |   226   (1)| 00:00:03 |
Predicate Information (identified by operation id):
   1 - access("T1"."C1"="T2"."C1")
   2 - filter("T2"."C1">=1 AND "T2"."C1"<=100000)
   4 - access("T1"."C1">=1 AND "T1"."C1"<=100000)
Statistics
          0 recursive calls
          0 db block gets
      37720 consistent gets
      25225 physical reads
          0 redo size
   82555058 bytes sent via SQL*Net to client
      73722 bytes received via SQL*Net from client
       6668 SQL*Net roundtrips to/from client
          0 sorts (memory)
          0 sorts (disk)
     100000 rows processed
STAT lines from the 10046 trace:
STAT #38 id=1 cnt=100000 pid=0 pos=1 obj=0 op='HASH JOIN (cr=37720 pr=25225 pw=0 time=69225424 us)'
STAT #38 id=2 cnt=100000 pid=1 pos=1 obj=48144 op='TABLE ACCESS FULL T2 (cr=12511 pr=12501 pw=0 time=13204971 us)'
STAT #38 id=3 cnt=100000 pid=1 pos=2 obj=48141 op='TABLE ACCESS BY INDEX ROWID T1 (cr=25209 pr=12724 pw=0 time=66504913 us)'
STAT #38 id=4 cnt=100000 pid=3 pos=1 obj=48142 op='INDEX RANGE SCAN INT_T1_C1 (cr=6876 pr=224 pw=0 time=604405 us)'
SQL> SELECT /*+ USE_NL(T1 T2) */
2    T1.C1,
3    T2.C2,
4    T1.C3
5 FROM
6    T1,
7    T2
8 WHERE
9    T1.C1=T2.C1
10    AND T1.C1 BETWEEN 1 AND 100000;
100000 rows selected.
Elapsed: 00:00:28.11
Execution Plan
Plan hash value: 1467726760
| Id | Operation                   | Name      | Rows | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |           | 99999 |    77M|   303K (1)| 01:00:43 |
|   1 | TABLE ACCESS BY INDEX ROWID| T1        |     1 |   806 |     3   (0)| 00:00:01 |
|   2 |   NESTED LOOPS              |           | 99999 |    77M|   303K (1)| 01:00:43 |
|* 3 |    TABLE ACCESS FULL        | T2        |   100K|   976K| 3435   (1)| 00:00:42 |
|* 4 |    INDEX RANGE SCAN         | INT_T1_C1 |     1 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - filter("T2"."C1">=1 AND "T2"."C1"<=100000)
   4 - access("T1"."C1"="T2"."C1")
       filter("T1"."C1"<=100000 AND "T1"."C1">=1)
Statistics
          0 recursive calls
          0 db block gets
     250218 consistent gets
      25225 physical reads
          0 redo size
   82555058 bytes sent via SQL*Net to client
      73722 bytes received via SQL*Net from client
       6668 SQL*Net roundtrips to/from client
          0 sorts (memory)
          0 sorts (disk)
     100000 rows processed
STAT lines from the 10046 trace:
STAT #26 id=1 cnt=100000 pid=0 pos=1 obj=48141 op='TABLE ACCESS BY INDEX ROWID T1 (cr=250218 pr=25225 pw=0 time=80712592 us)'
STAT #26 id=2 cnt=200001 pid=1 pos=1 obj=0 op='NESTED LOOPS (cr=231885 pr=12725 pw=0 time=4601151 us)'
STAT #26 id=3 cnt=100000 pid=2 pos=1 obj=48144 op='TABLE ACCESS FULL T2 (cr=18344 pr=12501 pw=0 time=17704737 us)'
STAT #26 id=4 cnt=100000 pid=2 pos=2 obj=48142 op='INDEX RANGE SCAN INT_T1_C1 (cr=213541 pr=224 pw=0 time=2683089 us)'
SQL> SELECT /*+ USE_NL(T1 T2) FIND_ME */
2    T1.C1,
3    T2.C2,
4    T1.C3
5 FROM
6    T1,
7    T2
8 WHERE
9    T1.C1=T2.C1
10    AND T1.C1 BETWEEN 1 AND 100000;
100000 rows selected.
Elapsed: 00:00:17.81
Execution Plan
Plan hash value: 1467726760
| Id | Operation                   | Name      | Rows | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |           | 99999 |    77M|   303K (1)| 01:00:43 |
|   1 | TABLE ACCESS BY INDEX ROWID| T1        |     1 |   806 |     3   (0)| 00:00:01 |
|   2 |   NESTED LOOPS              |           | 99999 |    77M|   303K (1)| 01:00:43 |
|* 3 |    TABLE ACCESS FULL        | T2        |   100K|   976K| 3435   (1)| 00:00:42 |
|* 4 |    INDEX RANGE SCAN         | INT_T1_C1 |     1 |       |     2   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   3 - filter("T2"."C1">=1 AND "T2"."C1"<=100000)
   4 - access("T1"."C1"="T2"."C1")
       filter("T1"."C1"<=100000 AND "T1"."C1">=1)
Statistics
          0 recursive calls
          0 db block gets
     250218 consistent gets
          0 physical reads
          0 redo size
   82555058 bytes sent via SQL*Net to client
      73722 bytes received via SQL*Net from client
       6668 SQL*Net roundtrips to/from client
          0 sorts (memory)
          0 sorts (disk)
     100000 rows processed
STAT lines from the 10046 trace:
STAT #36 id=1 cnt=100000 pid=0 pos=1 obj=48141 op='TABLE ACCESS BY INDEX ROWID T1 (cr=250218 pr=0 pw=0 time=6000438 us)'
STAT #36 id=2 cnt=200001 pid=1 pos=1 obj=0 op='NESTED LOOPS (cr=231885 pr=0 pw=0 time=2401295 us)'
STAT #36 id=3 cnt=100000 pid=2 pos=1 obj=48144 op='TABLE ACCESS FULL T2 (cr=18344 pr=0 pw=0 time=1400071 us)'
STAT #36 id=4 cnt=100000 pid=2 pos=2 obj=48142 op='INDEX RANGE SCAN INT_T1_C1 (cr=213541 pr=0 pw=0 time=2435627 us)'So, what does the above mean? Why would a forced execution plan change the number of consistent gets? Why would not flushing the buffer cache cause the PR statistic to drop to 0, yet not change the CR statistic?
Charles Hooper
IT Manager/Oracle DBA
K&M Machine-Fabricating, Inc.

How can I read the image from database into form

hi everyone
if I have table like this
create table test
(id number(10),
pic long raw);
in this table record
in form If I have block non database
how can i read the image into non database item
I try with this code but no good
select pic into :photo from test where id=5;
photo its non database item kind image
how can i invoke image into item
thanks alooooooot

Hello,
You can't select image with select query. You need to create another database block from test table which has item binds with database pic column. Now in where condition in test block's property you have to write id=5. Change test block's properties UPDATE ALLOWED, INSERT ALLOWED, DELETE ALLOWED to NO. Only Query allowed property set to YES.
Now in your particular trigger write the code
GO_BLOCK('test');
Execute_Query;
please mark if it help you or correct
Regards,
Danish

Single block read for Sort

I’m creating a temp table with following SQL. Table temp2 is 50GB in size with 120M rows. I noticed during SORT operation , it’s performing lots of single block reads on Temp tablespace. Why it’s doing single block reads on temp??
Also, Is there any other way to make this quicker?
I'm on 10gR2 with Auto PGA/SGA. PGA Aggregate is 5GB and SGA is 12GB.
create table t_temp nologging parallel(degree 4 )
as
select /*+ parallel(t1,4) */ * from temp2 partition(P2008) t1
order by custkey
    SID Wait State    EVENT                               P1                         P2              P3                    SEQ# % TotTime TEventTime(ms) DistEvnts Avgtime(ms)/Evnt
   2165 WAITING       direct path read temp               file number= 5001          first dba=      block cnt= 1                     .02           .253          1             .253
                                                                                     88791
   2165 WAITING       direct path read temp               file number= 5001          first dba=      block cnt= 1                     .02           .253          1             .253
                                                                                     412771
   2165 WAITING       direct path read temp               file number= 5001          first dba=      block cnt= 1                     .02           .253          1             .253
                                                                                     421465
   2165 WAITING       direct path read temp               file number= 5001          first dba=      block cnt= 1                     .02           .253          1             .253
                                                                                     691141
   2165 WAITING       direct path read temp               file number= 5001          first dba=      block cnt= 1                     .02           .253          1             .253
                                                                                     1295425
Here is Temp table space properties…
Name                    INITK    NEXTK MINEX MAXEX PCTI CNTS      EXMGMT     ALOCTYPE SEGMGM STATUS
TEMP                     1024     1024     1          0 TEMPORARY LOCAL      UNIFORM   MANUAL ONLINEEdited by: user4529833 on Feb 5, 2009 9:05 AM

user4529833 wrote:
Thanks Jonathan.
I will do that and post the results... One questions though, how to correct this unexpected side-effect of asynchronous reads... ( Assuming it's the case )
If the assumption is correct, then it's not really something that needs to be corrected. Because of the overlapping of I/Os, there may be lots of I/Os that never make it into wait time, so a single process could be going:
give me 12 blocks
give me 12 blocks
give me 12 blocks
give me 12 blocks
give me a block
of which you see only the "give me a block".
That's why the 10033 is an important diagnostic - it will help you decide whether or not "everything" is a single block read, or whether the single block reads are a very tiny fraction of the total I/O that also happens to become visible for some reason.
how does oracle perform sort for SQLs I have given..
1) read & sort the data for custid coulmn with rowids and dump it temp (quite likely)
2) read the rowids from temp and start loading row to temp table..I'd have to check the execution to give you a definite answer but the basic strategy would be:
<ul>
Two sets of PX slaves.
Set one reads a chunk of the source table picking up all the relevant columns for the result table.
Because of the ORDER BY clause layer one does a range-based distribution of data to the second set of slaves - along the lines of "slave 1 gets custid less than 1M", "slave 2 gets custid between 1M and 2M" and so on. Because entire rows are passed between the two sets of slaves, you can see a LOT of PQ traffic going on. (People sometimes increase the "PX message size" parameter because of the volume of traffic and the size of the rows).
Set two slaves accumulate, and sort the incoming data sets. Each sorted row carries ALL the columns, there is no "sort the custkey/rowids only" strategy. The sorting may spill to temp - given your "select *", the amount of temp space used may actually be greater than the size of the source table.
Each set two slave creates its own "private table" with its results.
The query coordinator links the "private table" metadata in the data dictionay in the right order to create the final table.
</ul>
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
http://www.jlcomp.demon.co.uk
"Science is more than a body of knowledge; it is a way of thinking"
Carl Sagan
To post code, statspack/AWR report, execution plans or trace files, start and end the section with the tag {noformat}{noformat} (lowercase, curly brackets, no spaces) so that the text appears in fixed format.

How do I read uncommitted records of another session

HI
How do I read uncommitted records from another session in Oracle
Thanks
Ashwin

Ashwin,
Oh I might as well add my thoughts on this as it is kinda an important subject. I'm sort of a different kind of programmer where there have been several times I have wanted to see the changed results that another session is doing. I agree that the rest of the sessions should not be able to SEE the changes until they are committed. HOWEVER, as some examples show below, I don't see why a custom package can't be developed to allow a different session to be able to query the changed data for a particular SID, or even begin issuing DML commands of their own on this changed data to help change it further.
In reality, there is no reason why a single (and I emphasize single) session is the only one who can see the changed data. It's the other user sessions that really should wait until all the necessary changes are complete before others can see the changed information. There isn't much point in me trying to explain this in great detail, but overall YES other users should wait until the changes are committed, but NO I do think I should be able to connect to another SIDs sessions and be able to view/perform operations on this changed data.
There are more than enough examples of why others should wait before changes are committed (which I agree), but as some of these real life examples point out, there is a need to let more than a single session perform such changes, each of which need to see each other's changed data.
No I am no expert on transaction processing (Andrew knows more for sure), but I've done enough to know as a developer I have a number of times wanted to be able to access/monitor another sessions uncommitted data. I think some records might be viewable using committed AUTONOMOUS_TRANSACTION records, but I've yet to really try it out much.
Examples where one wants to see others uncommitted data
(+) Say I want to change a hundred / thousand / million tables for a task. Each of these tables require joins/lookups to see each others data (including the changed records). I'm not going to want to do this with one session. Why can't 10 or a 100 sessions be created to allow complex DML operations to be performed to accomplish a given task. It's like thinking of there is 1 session doing this task, but on the oracle side, there are a unlimited number of SIDs processing (100 sqlplus scripts running), all being able to see the uncommitted data. Once everything is done, then the rest of the sessions can see this information.
(+) A process is running and is changing records, and I want to monitor its progress or even assist it it's getting behind.
(+) A real life example I just remembered was a soccer game done overseas where a time zone problem allowed people to still be able to place bets even after the game was over (they still paid people even though they knew the final score which was nice of them). In this case yes a process should have been stopped from allowing bets/inserts to continued, but having a second (or more) processes being able to KNOW for certain that another session is inserting data when it shouldn't be, this is can be stopped. And yes once again you can say one can monitor the sga for the appearance of inserts and stop it and yes remove the insert privs from that user after a certain time (come to think of it that wouldn't have worked since the time was off). But yes a whole slue of other things could have been done to stop this process from recording this information. However, NOT being able to select data from that table where the inserts were going into, until its toooooooooo late, is a real problem. Being able to see uncommitted information is very important if it needs to be stopped.
Overall I do think there should be something considered to allow to a session to be able to see what data another session is doing, but it's more on the side from administration of the data and the performance required to get a task done, even if it means sharing uncommitted changes between sessions.
Tyler D.

How to keep Reader from being installed if running Pro? Running Acrobat X Pro. Occaisionally, a website with a PDF triggers Reader to be installed on my system. NEVER want reader (that's why i bought Pro). How do i keep Reader from EVER being installed?

Running Acrobat X Pro.
Occasionally, a website with a PDF triggers Reader to be installed on my system.
NEVER want reader (that's why i bought Pro).
How do i keep Reader from EVER being installed?

I don't get a pop-up. Something triggers the Reader installer to be loaded on my system. This occurs without my knowledge. Then, the next time i go to open a PDF, instead of it opening immediately with Acrobat Pro, I get the first screen of the installation stuff for Reader (Accept/Decline Terms of use, etc.). At this point, i get out of the install, go into control panel and uninstall Reader (which is listed as a loaded program, even though i haven't finished the install). After i uninstall reader and open a PDF, it opens correctly with Acrobat Pro. This happens once or twice a month at least. Is it possible that the Adobe Acrobat Update Service task loads Reader on my system at some interval? How do i block Reader form EVER being installed or loaded?

How can I read email from my aol account from my iphone and keep it as uread on my computer at home. It automatically goes to read mail on my computer. On my computer if I want to keep an email to answer later I can mark it "Keep as New".

How can I read my email from my aol account from my iphone and keep it as "unread" on my computer at home? At home I can read an email and if I want to get back to it at a later date I can mark it as "keep as new". I tend to forget it if it goes to "read" mail. Right now, when I read an email from my phone it goes automatically to "read" mail.

On the iPad, using the mail app, there is no way to do what you are asking without tapping the flag icon and marking the item as unread. Have you tried the OWA app for the iPad? It may have that functionality, but I haven't tested it as you need an Office 365 subscription with Exchange support to use the app.

How to download / read text attachment in Sender Mail Adapter IN XI

Hi
I would like to know how to download / read text attachment in sender mail Adapter & sent same attachment to target system using file adapter.
Please help how to design / resolve this concept.
Regards
DSR

I would like to know how to download / read text attachment in sender mail Adapter & sent same
attachment to target system using file adapter.
Take help from this blog:
/people/michal.krawczyk2/blog/2005/12/18/xi-sender-mail-adapter--payloadswapbean--step-by-step
From the blog:
However in most cases
our message will not be a part of the e-mail's payload but will be sent as a file attachment.
Can XI's mail adapter handle such scenarios? Sure it can but with a little help
from the PayloadSwapBean adapter module
Once your message (attachment) is read by the sender CC, you can perform the basic mapping requirement (if any) to convert the mail message fromat to the file format.....configure a receiver FILE CC and send the message...this should be the design...
Regards,
Abhishek.

I am confused about something. How do I read a book on my MacBook Pro? I can't find the iBook app anywhere, which is what I use on my iPad. The book I want to read is in my iTunes but I can't click on it. My iBook library does not show up in iTunes.

I am confused about something. How do I read a book on my MacBook Pro? I can't find the iBook app anywhere, which is what I use on my iPad. The book I want to read is in my iTunes but I can't click on it. Some of my iBooks show up in my iTunes but they are "grayed" out. The only books that respond in iTunes are audiobooks and that's not what I'm looking for. Is this a stupid question?

Nevermind - I answered my own question, which is I CAN"T READ ANY BOOKS I purchased in iBooks on my MacBook Pro. If I want to read on my mac I have to use Kindle or Nook. Which means any book I've already purchased through iBooks has to be read on my iPad. Kind of a drag because there are times when it's more convenient for me to read while I'm sitting with my Mac.

How to block reading in "read commited"?

Similar Messages

Maybe you are looking for