Update taking time in sql - sql

Question

I have one query regarding update
update stg set flg = 'Y' where exists (
select 1
from err
where err.pid = stg.pid
and err.imp_id = stg.imp_id
and err.error_id = 999 )
and pid = 123 ;
so staging table should update its records when entry is there in err table
for 999 error_id. Err
table has 3mn records for 999.Stg has total 13mn records for 123 pid. Both
tables are partition
on pid. Err has index on all 3 cols. This is running so slow.
Please suggest any better approach for doing this. I have to do this in
plsql block.ALso let me know if more details are required. This job will run daily. No foreign keys .

Answer 1

You may try to transform your update into an inline view update, look for an example here: https://oracle-base.com/articles/misc/updates-based-on-queries
So I tried to reproduce a test case with the information you did provide:
Connected to:
Oracle Database 19c Enterprise Edition Release 19.0.0.0.0 - Production
Version 19.3.0.0.0
SQL> create table err (pid number not null, imp_id number not null, error_id number not null);
Table created.
SQL> create table stg (pid number not null, imp_id number not null, flg char(1) not null );
Table created.
SQL> insert into stg select 123, level, 'N' from dual connect by level <= 13000000;
13000000 rows created.
SQL> commit;
Commit complete.
SQL> insert into err select 123, level, 999 from dual connect by level <= 3000000;
3000000 rows created.
SQL> commit;
Commit complete.
SQL> set timing on
SQL> update stg set flg = 'Y' where exists (
select 1
from err
where err.pid = stg.pid
and err.imp_id = stg.imp_id
and err.error_id = 999 )
and pid = 123;
3000000 rows updated.
Elapsed: 00:00:13.63
SQL> rollback;
Rollback complete.
Elapsed: 00:00:09.71
SQL> create unique index udx_err on err (pid, imp_id, error_id);
Index created.
Elapsed: 00:00:06.21
SQL> create index idx_stg on stg (pid, imp_id);
Index created.
Elapsed: 00:00:23.21
SQL> exec dbms_stats.gather_table_stats( user, 'ERR' );
PL/SQL procedure successfully completed.
Elapsed: 00:00:02.11
SQL> exec dbms_stats.gather_table_stats( user, 'STG' );
PL/SQL procedure successfully completed.
Elapsed: 00:00:09.44
SQL> update stg set flg = 'Y' where exists (
select 1
from err
where err.pid = stg.pid
and err.imp_id = stg.imp_id
and err.error_id = 999 )
and pid = 123;
3000000 rows updated.
Elapsed: 00:00:13.84
SQL> explain plan for update stg set flg = 'Y' where exists (
select 1
from err
where err.pid = stg.pid
and err.imp_id = stg.imp_id
and err.error_id = 999 )
and pid = 123;
Explained.
Elapsed: 00:00:00.01
SQL> SELECT * FROM TABLE(DBMS_XPLAN.display(format => 'ALL'));
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------
Plan hash value: 728722196
--------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |
--------------------------------------------------------------------------------------
| 0 | UPDATE STATEMENT | | 3002K| 74M| | 29231 (1)| 00:00:02 |
| 1 | UPDATE | STG | | | | | |
|* 2 | HASH JOIN RIGHT SEMI| | 3002K| 74M| 74M| 29231 (1)| 00:00:02 |
|* 3 | TABLE ACCESS FULL | ERR | 2999K| 40M| | 2146 (2)| 00:00:01 |
|* 4 | TABLE ACCESS FULL | STG | 12M| 148M| | 8547 (2)| 00:00:01 |
--------------------------------------------------------------------------------------
Query Block Name / Object Alias (identified by operation id):
-------------------------------------------------------------
1 - SEL$3FF8579E
3 - SEL$3FF8579E / ERR#SEL$1
4 - SEL$3FF8579E / STG#UPD$1
Predicate Information (identified by operation id):
---------------------------------------------------
2 - access("ERR"."PID"="STG"."PID" AND "ERR"."IMP_ID"="STG"."IMP_ID")
3 - filter("ERR"."ERROR_ID"=999 AND "ERR"."PID"=123)
4 - filter("PID"=123)
Column Projection Information (identified by operation id):
-----------------------------------------------------------
2 - (#keys=2; upd=4; cmp=1,2) "STG"."PID"[NUMBER,22], "STG"."IMP_ID"[NUMBER,22], "STG".ROWID[ROWID,10], "FLG"[CHARACTER,1]
3 - (rowset=256) "ERR"."PID"[NUMBER,22], "ERR"."IMP_ID"[NUMBER,22], "ERR"."ERROR_ID"[NUMBER,22]
4 - "STG".ROWID[ROWID,10], "PID"[NUMBER,22], "STG"."IMP_ID"[NUMBER,22], "FLG"[CHARACTER,1]
35 rows selected.
Elapsed: 00:00:00.08
Using inline view now:
SQL> update (select stg.flg
from stg, err
where err.pid = stg.pid
and err.imp_id = stg.imp_id
and err.error_id = 999
and stg.pid = 123
)
set flg = 'Y';
3000000 rows updated.
Elapsed: 00:00:05.69
SQL>
And it runs faster...
My advices:
- check your statistics
- check these indexes, are they well defined?
- check your columns data types
- check if columns should be not nullable
Your execution plan should not use indexes here as it will run forever...