Question about cardinality (lines) to explain the plan

Similar Questions

• Explain the plans differ as the parameter value changes

  Hi all
  
  My colleague posted a similar question a few days before. Happened because of some bad index. But now we are in a strange situation.
  
  DB:
  

  SQL> select * from v$version;
  
  BANNER
  ----------------------------------------------------------------
  Oracle Database 10g Enterprise Edition Release 10.2.0.1.0 - Prod
  PL/SQL Release 10.2.0.1.0 - Production
  CORE    10.2.0.1.0      Production
  TNS for 32-bit Windows: Version 10.2.0.1.0 - Production
  NLSRTL Version 10.2.0.1.0 - Production

  We use the query below and was working fine until 13.
  

  SQL> explain plan for
    2  SELECT *
    3    FROM gacc_dtl_v1  acc,
    4         gcus_dtl_v1  cus,
    5         gtxn_dtl_v1  txn
    6   WHERE txn.customer_id = cus.customer_number(+)
    7   AND txn.batch_id = cus.batch_id(+)
    8   AND txn.account_number = acc.id
    9   AND acc.batch_id = '130609'
   10   AND cus.batch_id(+) = '130609'
   11   AND txn.batch_id = '130609' AND cus.target IN ('30');
  
  Explained.
  
  SQL> select * from table(dbms_xplan.display);
  
  PLAN_TABLE_OUTPUT
  -------------------------------------------------------------------------------------------------------------
  Plan hash value: 566819363
  
  -------------------------------------------------------------------------------------------------------------
  | Id  | Operation                     | Name                | Rows  | Bytes |TempSpc| Cost (%CPU)| Time     |
  -------------------------------------------------------------------------------------------------------------
  |   0 | SELECT STATEMENT              |                     |     1 |   947 |       | 16963   (1)| 00:03:24 |
  |   1 |  NESTED LOOPS                 |                     |     1 |   947 |       | 16963   (1)| 00:03:24 |
  |*  2 |   HASH JOIN                   |                     |    41 | 26322 |  9136K| 16799   (1)| 00:03:22 |
  |*  3 |    TABLE ACCESS BY INDEX ROWID| GTXN_DTL_V1         | 31055 |  8764K|       |  2430   (1)| 00:00:30 |
  |*  4 |     INDEX RANGE SCAN          | GTXN_V1_BATCHID_NDX | 60524 |       |       |   156   (2)| 00:00:02 |
  |*  5 |    TABLE ACCESS BY INDEX ROWID| GCUS_DTL_V1         |   176K|    59M|       | 10869   (1)| 00:02:11 |
  |*  6 |     INDEX RANGE SCAN          | IDX_CUS2_V1         |   198K|       |       |   527   (2)| 00:00:07 |
  |   7 |   TABLE ACCESS BY INDEX ROWID | GACC_DTL_V1         |     1 |   305 |       |     4   (0)| 00:00:01 |
  |*  8 |    INDEX RANGE SCAN           | GACC_DTL_V1_IDX     |     1 |       |       |     3   (0)| 00:00:01 |
  -------------------------------------------------------------------------------------------------------------
  
  Predicate Information (identified by operation id):
  ---------------------------------------------------
  
     2 - access("TXN"."CUSTOMER_ID"="CUS"."CUSTOMER_NUMBER" AND "TXN"."BATCH_ID"="CUS"."BATCH_ID")
     3 - filter("TXN"."CUSTOMER_ID" IS NOT NULL)
     4 - access("TXN"."BATCH_ID"='130609')
     5 - filter("CUS"."TARGET"='30')
     6 - access("CUS"."BATCH_ID"='130609')
     8 - access("TXN"."ACCOUNT_NUMBER"="ACC"."ID" AND "ACC"."BATCH_ID"='130609')
         filter(SUBSTR("TXN"."ACCOUNT_NUMBER",1,3)=SUBSTR("ACC"."ID",1,3))
  
  26 rows selected.

  It shows a hash join and nested with cost 16963 loops and gives the result in 2-3 seconds. It gives the same plan to explain even now if we use batch_id = '130609'
  
  Now all of a sudden from yesterday it gives different explain the plan below. Only difference in the query below is the value of batch_id

  SQL> explain plan for
    2  SELECT *
    3    FROM gacc_dtl_v1  acc,
    4         gcus_dtl_v1  cus,
    5         gtxn_dtl_v1  txn
    6   WHERE txn.customer_id = cus.customer_number(+)
    7   AND txn.batch_id = cus.batch_id(+)
    8   AND txn.account_number = acc.id
    9   AND acc.batch_id = '150609'
   10   AND cus.batch_id(+) = '150609'
   11   AND txn.batch_id = '150609' AND cus.target IN ('30');
  
  Explained.
  
  SQL> select * from table(dbms_xplan.display);
  
  PLAN_TABLE_OUTPUT
  ------------------------------------------------------------------------------------------------------------
  Plan hash value: 773603995
  
  --------------------------------------------------------------------------------------------------------
  | Id  | Operation                     | Name                   | Rows  | Bytes | Cost (%CPU)| Time     |
  --------------------------------------------------------------------------------------------------------
  |   0 | SELECT STATEMENT              |                        |     1 |   947 |    77   (0)| 00:00:01 |
  |   1 |  NESTED LOOPS                 |                        |     1 |   947 |    77   (0)| 00:00:01 |
  |   2 |   NESTED LOOPS                |                        |     1 |   594 |    73   (0)| 00:00:01 |
  |   3 |    TABLE ACCESS BY INDEX ROWID| GACC_DTL_V1            |     1 |   305 |     4   (0)| 00:00:01 |
  |*  4 |     INDEX RANGE SCAN          | GACC_DTL_BATCH_ID_INDX |     1 |       |     3   (0)| 00:00:01 |
  |*  5 |    TABLE ACCESS BY INDEX ROWID| GTXN_DTL_V1            |     1 |   289 |    69   (0)| 00:00:01 |
  |*  6 |     INDEX RANGE SCAN          | IDX_TXN2_V1            |   125 |       |    12   (0)| 00:00:01 |
  |*  7 |   TABLE ACCESS BY INDEX ROWID | GCUS_DTL_V1            |     1 |   353 |     4   (0)| 00:00:01 |
  |*  8 |    INDEX RANGE SCAN           | IDX_CUS3_V1            |     1 |       |     3   (0)| 00:00:01 |
  --------------------------------------------------------------------------------------------------------
  
  Predicate Information (identified by operation id):
  ---------------------------------------------------
  
     4 - access("ACC"."BATCH_ID"='150609')
     5 - filter("TXN"."CUSTOMER_ID" IS NOT NULL AND "TXN"."BATCH_ID"='150609')
     6 - access("TXN"."ACCOUNT_NUMBER"="ACC"."ID")
         filter(SUBSTR("TXN"."ACCOUNT_NUMBER",1,3)=SUBSTR("ACC"."ID",1,3))
     7 - filter("CUS"."TARGET"='30')
     8 - access("CUS"."BATCH_ID"='150609' AND "TXN"."CUSTOMER_ID"="CUS"."CUSTOMER_NUMBER")
         filter("TXN"."BATCH_ID"="CUS"."BATCH_ID")
  
  26 rows selected.

  It shows two loops nested with cost 77, but works for hours. Very very slow.. No idea what's going on...
  

   select i.table_name,i.index_name,index_type,c.column_name,c.column_position,e.column_expression
     from all_indexes i, all_ind_columns c,all_ind_expressions e
     where c.index_name = i.index_name
     and e.index_name(+) = i.index_name
     and i.table_name in ('GCUS_DTL_V1','GACC_DTL_V1','GTXN_DTL_V')
     order by 1,2,4
  
  TABLE_NAME     INDEX_NAME          INDEX_TYPE          COLUMN_NAME     COLUMN_POSITION     COLUMN_EXPRESSION
  --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  GACC_DTL_V1     GACC_DTL_BATCH_ID_INDX     NORMAL               BATCH_ID          1     
  GACC_DTL_V1     GACC_DTL_V1_IDX          NORMAL               BATCH_ID          2     
  GACC_DTL_V1     GACC_DTL_V1_IDX          NORMAL               ID               1     
  GACC_DTL_V1     GACC_DTL_V1_IDX2     FUNCTION-BASED NORMAL     SYS_NC00101$          1     SUBSTR("ID",1,3)
  GACC_DTL_V1     IDX_ACC1_V1          NORMAL               CATEGORY          1     
  GACC_DTL_V1     IDX_ACC3_V1          FUNCTION-BASED NORMAL     SYS_NC00099$          1     "CUSTOMER_NUMBER"||'.'||"LIMIT_REF"
  GACC_DTL_V1     IDX_ACC4_V1          FUNCTION-BASED NORMAL     SYS_NC00100$          1     "CUSTOMER_NUMBER"||'.000'||"LIMIT_REF"
  GACC_DTL_V1     IDX_ACC5_V1          NORMAL               POSTING_RESTRICT     1     
  GACC_DTL_V1     IDX_CUS5_V1          NORMAL               CUSTOMER_NUMBER          1     
  GACC_DTL_V1     IDX_CUS6_V1          NORMAL               LIMIT_REF          1     
  GCUS_DTL_V1     GCUS_DTL_V1_IDX1     NORMAL               CUSTOMER_NUMBER          1     
  GCUS_DTL_V1     IDX_CUS2_V1          NORMAL               BATCH_ID          1     
  GCUS_DTL_V1     IDX_CUS3_V1          NORMAL               BATCH_ID          1     
  GCUS_DTL_V1     IDX_CUS3_V1          NORMAL               CUSTOMER_NUMBER          2     
  GCUS_DTL_V1     IDX_CUS3_V1          NORMAL               INDUSTRY          4     
  GCUS_DTL_V1     IDX_CUS3_V1          NORMAL               SECTOR               3     
  GCUS_DTL_V1     IDX_CUS4_V1          FUNCTION-BASED NORMAL     SYS_NC00078$          1     SUBSTR("DATE_STAMP",1,6)

  We are also do not understand why the filter (SUBSTR ("TXN". ""»(, 1, 3) ACCOUNT_NUMBER = SUBSTR ("VAC". " ID", 1, 3)) is used in both queries.
  
  All tables are analyzed today.
  
  Please share your thoughts on this.
  
  Thanks in advance,
  Jac

  Jac says:

  L     H     NUM_BUCKETS     LAST_ANALYZED     SAMPLE_SIZE     HISTOGRAM
  ------------------------------------------------------------------------------------------------------------------------------------------------
  010109     311208     235     13/Jun/2009     5,343     FREQUENCY
  

  You have a histogram of frequencies on the BATCH_ID column missing at least 2 values according to your index statistics (235 buckets vs 237 separate keys).

  If the value that you use in the query is missing then this could be the explanation for the estimation of cardinality bad (since you're on pre - 10.2.0.4. In 10.2.0.4 that this behavior changes).

  The size of the sample of 5 300 lines is also very low, given the 57,000,000 lines according to the index statistics.

  You have two options (which can be combined):

  -Increase the size of the sample using a parameter explicitly estimate_percent, for example at least 10 percent

  exec DBMS_STATS. GATHER_TABLE_STATS (null, 'GACC_DTL_V1', estimate_percent-online 10, method_opt => 'FOR COLUMNS SIZE 254 BATCH_ID,' waterfall-online fake)

  -Get rid of the histogram

  exec DBMS_STATS. GATHER_TABLE_STATS (null, 'GACC_DTL_V1', method_opt-online 'FOR BATCH_ID COLUMNS SIZE 1', cascade-online fake)

  Note: Is there a particular reason why you store numbers in varchar columns? This might be the reason why Oracle believes that it must generate a histogram using the AUTO SIZE option.

  I tend to promote to remove from the histogram, but you must first verify the data if the BATCH_ID values are spread out and the histogram is reasonable:

  select
          batch_id
        , count(*)
  from
          gacc_dtl_v1
  group by
          batch_id;
  

  No constarints are at the DB level. All are processed Application level.

  Have you checked this in DBA/ALL/USER_CONSTRAINTS?

  It's also a good idea to have constraints at the level of the DB. It keeps your data consistent and quite often helps the optimizer. It allows even 10.2 and later to make things like the elimination of a join table that can make a huge difference in performance.

  Kind regards
  Randolf

  Oracle related blog stuff:
  http://Oracle-Randolf.blogspot.com/

  SQLTools ++ for Oracle (Open source Oracle GUI for Windows):
  http://www.sqltools-plusplus.org:7676 /.
  http://sourceforge.NET/projects/SQLT-pp/

  Published by: Randolf Geist on June 16, 2009 11:48

  Comment added constraints

• a sql id have more than explain the plan so I'd like to come up with plan used at runtime.

  Hi all

  
  

  In a sql oracle11g having id several explain the plan so I'd like to come up with plan used at runtime.

  For example SQlID:-8yczg5zav14vt have 5 sql plan and I want to check that we execute at any time, please let me how I can check by sql queries

  
  

  Concerning

  Ranjeet
  

  RanjeetSohale wrote:

  I am ok for that but active only plans is right, but in accordance with the foregoing, two data line table in the table SQL V$ so both active and plan cost is high and the other is low.

  Yes that's right cost is different, he means here another effective plan for oracle will analyze the declaration again and stored as new slider of the child. Especially since I explained that.

• I plan to change the plan of creative photography of cloud. I currently use 5.7 LR with photoshop elements as my level of pixel Editor.  The question I have is if with the plan CC I can choose to download photoshop CC only for now. I'd do it

  I plan to change the plan of creative photography of cloud. I currently use 5.7 LR with photoshop elements as my level of pixel Editor.  The question I have is if with the plan CC I can choose to download photoshop CC only for now. I prefer wait to download LR CC if possible, but didn't know if I could download one without the other. Thanks for your time.

  Hello

  Yes, you could download Photoshop cc only as well.

  Concerning

  Gilles marneffe

• explain the plan and so on

  Hello world
  
  
  could someone please provide me with details of explained the plan and I would appreciate some related details explain plan, trace and tkprof
  
  Thank you
  Shareef

  Hello

  PLAN of EXPLAINING is an Oracle utility that analyzes a statement and shows the expected execution plan. It may be different from the implementation plan real for a number of reasons.

  Extended SQL trace (10046 event) is a way to gather advanced diagnostic information. There are different levels, at levels 8 and 12 provide the greatest level of detail (you will be able to see the events of waiting and the bind variable values).

  TKProf is a utility that processes the raw trace files and makes shaped the output in a more readable way. However, some experts prefer to work with trace files "gross" - they are not so difficult to read, especially if you get some practice.

  There are other events of tracing: 10053 allows you to see what is happening inside the Oracle optimizer, 10104 allows to see the stats of hash etc join.

  Best regards
  Nikolai

• difference between the execution plan and explain the plan?

  What is the difference between the execution plan & explain the plan?

  an execution plan is the actual steps that oracle will pass by when it executes a query.

  explain plan is a tool that is used to generate the steps of an execution plan for a query.

• explain the plan against the eve...

  I have a DDL which takes a long time to run on the primary
  would like to know how long it takes if I switch to sleep mode and run education instead of the primary current standby... is possible to run explain sleep agasint plan without having to go through the whole process of switch?
  I tried but the statement explain plan is a failure... Statement requires write Victorian for table plan and may not write in the plan table because db is not open... :) perhapts a new feature to implement in oracle 12g? :)
  :)
  
  v11.2.0.2
  physical standby

  be nice if we could run explains the plan for the system of relief through the primary system :) maybe in 12g?

  do not know now. :)

• Generation to EXPLAIN the PLAN on a database that is open in READ ONLY mode

  Hello
  
  I use the Oracle 10.2.0.3 version.
  
  If my database is opened in READ ONLY mode, means that no insert/update/delete operations are allowed here.
  
  During the generation of the PLAN to EXPLAIN this, the PLAN_TABLE registrations for any SQL. But my database is opened in READ ONLY mode, means that no inserts can happen.
  
  So, how can I generate EXPLAIN PLAN for my SQL in this State?
  
  Thanks in advance.
  
  
  
  Best regards
  oratest

  oratest wrote:
  I use the Oracle 10.2.0.3 version.

  If my database is opened in READ ONLY mode, means that no insert/update/delete operations are allowed here.

  During the generation of the PLAN to EXPLAIN this, the PLAN_TABLE registrations for any SQL. But my database is opened in READ ONLY mode, means that no inserts can happen.

  So, how can I generate EXPLAIN PLAN for my SQL in this State?

  You can always do: 'explain the plan in some_table@remote_database' to avoid inclusion in the local database. Unfortunately 10g added an extraction of the sequence of the code "explain plan", and that's where the call fails if you have tried this distant approach on your version.

  Here's an idea that I have not tested. If you configure a link of data to your database from production to the database read-only, you could then do a "explain plan" in the database of production for the SQL statement by changing each object reference in the SQL statement to "object@readonlydatabase". In most cases this will allow the optimizer to recognize the statement as "entirely to distance" and get the optimizer on the readonly database to create the execution plan - which will be then written into the production database.

  Concerning
  Jonathan Lewis

• why sometimes the trace file does not explain the plan?

  Hello
  
  Sometimes, when I 'alter session set sql_trace = true' and run some querys, some of them don't see explained the plan in the trace file?
  
  I tried "alter system RAS shared_pool" before starting the trace, but not luck.
  
  (Oracle 10g R2)
  
  any ideas?
  
  Thnks
  Miguel
  
  Published by: jmmnunes on Apr 27, 2010 18:01

  some of them don't watch not explain the plan in the trace file?

  Log out of your session after you set the trace sql false?

  All cursors must be closed to have access to all the information from the row source in the trace file.

• explain the plan of a query with variables

  Trying to Explain plan at some sql code in sql * more. The query has a variable. How can I do this?
  
  I look to explain the plan and dbms_xplan but did not find anything with variables

  use sqlplus variable bind:

  SQL> --define variable
  SQL> var x varchar2
  
  SQL> -- notice the colon prefixing the variable
  SQL> explain plan for select * from customer where cid = :x;
  
  Explained.
  
  SQL> select * from table( dbms_xplan.display );
  
  PLAN_TABLE_OUTPUT
  -----------------------------------------------------------------------------------------------
  Plan hash value: 1709312366
  
  ----------------------------------------------------------------------------------------
  | Id  | Operation                   | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
  ----------------------------------------------------------------------------------------
  |   0 | SELECT STATEMENT            |          |     1 |    67 |     2   (0)| 00:00:01 |
  |   1 |  TABLE ACCESS BY INDEX ROWID| CUSTOMER |     1 |    67 |     2   (0)| 00:00:01 |
  |*  2 |   INDEX RANGE SCAN          | CID      |     1 |       |     1   (0)| 00:00:01 |
  ----------------------------------------------------------------------------------------
  
  Predicate Information (identified by operation id):
  ---------------------------------------------------
  
     2 - access("CID"=:X)
  
  14 rows selected.
  

  the variable should not be set to explain the request, because explain does not actually run.

  Published by: shoblock on November 6, 2008 16:51

• Reg: Explain the Plan-

  Hi Experts,
  
  I had a doubt about the Plan to explain it.
  
  Is there a relationship between the cost amount in the Plan to explain and the number of lines read by the query?
  In other words - if the number of lines read by the query increases, which will increase the cost also?
  
  Can someone please give me some advice on this?
  
  
  Thank you and best regards,
  Vanessa B.

  ranitB wrote:
  Hi Experts,

  I had a doubt about the Plan to explain it.

  Is there a relationship between the cost amount in the Plan to explain and the number of lines read by the query?
  In other words - if the number of lines read by the query increases, which will increase the cost also?

  Can someone please give me some advice on this?

  "The answer is 'not really' the cost is single block read disc of research time. But if the number of lines that are expected + coming in overall result are higher, the work expected by Oracle would be also higher. Thus, indirectly, the answer may be Yes. That said, the number of rows contained in the plan of the explain command (you should check the execution plan, explain plan is not the real plan) are the lines that the optimizer is expected to pick up and extracted from the number ofrows real may or may not match with her.

  Aman...

• Questions about reinstalling Windows XP after the end of Support

  Several members of the family of my friends that I sometimes provide technical assistance for are still under XP SP3 on their systems.  These systems are too outdated and underpowered to upgrade to a newer version of Windows, and it will probably be some time until they are able to upgrade to a more recent hardware capable of running most modern versions of Windows.

  If in the meantime, something happens to one of their computers requiring a reinstall of Windows after support for XP purposes, but before they can get a newer computer, I have a few questions about what will happen:

  1. Is there still a way to activate their existing copies of XP during installation process?
  2. While I understand that new patches, updates, patches & will be is no longer provided, existing patches will be available through the Microsoft Update service so that I would be able to get those computers to update state that they would be about to XP supported ends, or do I have to search elsewhere to try & find these updates?  (It would be nice if Microsoft put up a cumulative update that would be sort of a mega-patch for people in this situation to be able to get back to the most current state systems, but I would not be surprised if there wasn't such a thing).

  Anyone know the answers to these questions?

  Thank you

  You should be able to clean install and activate Windows XP.  You should be able to upgrade as well, even if the wise move would be to get the updates stored on portable media or integrated into a multimedia installation (or both) as soon as POSSIBLE if the plan is to continue using Windows XP for a while.

  My suggestion is to set up and to buy Windows 7 Professional renovated systems / lease from places like TigerDirect.com - as pricing is generally about the same as just a license for Windows 7 or a little professional, the hardware specifications of these machines are likely machines Windows XP current in use, using something like the PC Mover take care of Laplink and Microsoft to move files (or for the more than 25 USD the) applications that can be moved as well) to the new system of the former is possible, and Windows 7 will get updated until 2020.

  Recently research I've seen systems to Windows 7 Professional for as little as 150 USD with 4 GB of memory and dual core processors.  Not bad.

• Questions about close cache events and the defeat strategy "ALL".

  I use near cache, and try to determine what defeat strategy would be better for me. My cards before being highly volatile, I'm considering strongly using the strategy of "EVERYTHING" for at least some of my caches.
  
  This brings me some questions about the event cache stream closely when the strategy using ALL:
  
  (1) should I expect an event to be published from the back cover when an element is added? (i.e. putAll() called cache before)
  (2) should I expect an event to be published from the back cover when an item is deleted? (i.e. remove() called cache before)
  (3) should I expect an event to be published from the back cover when an element is expelled from the rear because of size constraints?
  (4) should I expect an event to appear from the back cover when an element is expelled from the rear due to the expiry?

  Hi Tom,

  near cache before cards do not store the entries inserted in the back cover, so if you have saved a listener on the front plane, you will get the notification to the listener on the plan before all the inserts in the back cover of other nodes.

  It's a little different for entries that have been asked by the local node through the near cache (not directly on the distributed back cache). If the cache almost didn't get call so far, then they will not go into the front plane, so it's the same as if inserted directly into the back cover. It is the starting of the near cache behavior. After the first EEG puts will be cached in the front plane. Once that puts is cached in the front plane, then the listener before card will see put through the cache close to the same node, but not of other nodes.

  Also, the cache near registers a listener lite, so it becomes not the input values changed, as the keys.

  A listener that you entered on the back cover will see each insertion, if you do not save a MapEventFilter or a MapEventTransformer custom with the listener. A MapEventFilter can filter inserts away (I think, the cache similar to that, too).

  Best regards

  Robert

• Explain the Plan in SQL Developer vs SQLPlus

  Hello world
  
  I have a small question. Why a plan explain blocking in SQL Developer but return immediately in sqlplus?
  
  For example, in SQL developer, I write my statement and use the F6 command to bring up the plan to explain it and just, it hangs and crashes. It seems he's trying to produce a plan to explain, but it worked for hours now.
  
  When I log into SQL Plus, I do a

  SQL> explain plan for <query>:
  SQL> select * from table(dbms_xplan.display);

  The whole process takes less than a minute to get the plan of the explain command.
  
  Is there something fundamentally different running through SQL Developer?
  
  Thank you!
  
  -Joe

  Joe,

  F6 in SQL Developer runs an AutoTrace, which is currently running the SQL and also shows schedules and terms of the explain command. If you only want the plan of the explain command, use F10.

  Kind regards
  Bob

• explain the plan using plan_hash_value and runtime

  Hello

  Please let me know the syntax to get the plan to explain using sql_id and plan_hash_value.

  Also let me know how to find out the same sql execution times in the previous months. I have the ID value of sql.

  Kind regards

  VN

  > Is it possible to check how long the sql finished in the previous months?

  Of AWR / StatsPack reports.

  Or use awrsqrpt.sql which can query the repository for a single SQL_ID AWR.

  However, you must have the AWR data kept for 4 months.  The default retention is 7 days.

  Hemant K Collette