For example DBA would like to move datafiles to a new file system, from smaller file systems, not realising that there are duplicated file names under the different file systems, and when moved to a file system, one overwrites the other… 

 select *  from ( select  tablespace_name, fullpath, filename, count(filename) over (partition by filename)  freq from  ( select tablespace_name, file_name fullpath, substr(file_name, instr(file_name,‘/’,-1)+1)  filename from dba_data_files ) )  where freq > 1

select * from xx order by a;

1
2
4
6
7
11
12
13
14
120
We to change the linenumber so it is sequential (and not sparse), simply do

select a, row_number() over (order by a) newlinenumber
from xx;

And this will return
        

A NEWLINENUMBER
1 1
2 2
4 3
6 4
7 5
11 6
12 7
13 8
14 9
120 10

select * from xx order by a;

1
2
4
6
7
11
12
13
14
120
We need to present the data as follows (insert blank line every 2 lines).
1
2
<blank line>
4
6
<blank line>
7
11
<blank line>
12
13
<blank line>
14
120
To illustrate this, I will use a staging table yy
create table yy (linenumber number, a number);

The technique is to simply use mod function as follows:
declare
cursor cur_i is select rownum linenumber, a from xx order by a;
v_mod2 number;
v_count number :=1;
begin
for i in cur_i
loop
 select mod(i.linenumber,2) into v_mod2 from dual;
 if v_mod2 = 0 then
   insert into yy values (v_count, a);
   v_count := v_count+1;
   insert into yy values (v_count,null);
   v_count := v_count+1;
 else
   insert into yy values (v_count, a);
   v_count := v_count+1;
 end if;
end loop;
commit;
end;
/
 
 
And the result will be :

select * from yy;

1 1
2 2
3 null
4 4
5 6
6 null
7 7
8 11
9 null
10 12
11 13
12 null
13 14
14 120
15 null

Here I am trying to automate the creation of commands to merge subpartitions within a table.

The command output would look like this

alter table xxx merge subpartitions a,b,c,d,e,f into sp_xx;

declare
cursor cur_part is select distinct partition_name from user_tab_subpartitions where table_name=’ACCOUNT_POSTING’;
v_sql varchar2(255);
begin
  for i in cur_part
  loop
   
    select ‘alter table account_posting merge subpartitions ‘||substr(max(sys_connect_by_path(subpartition_name,’,')),2)||’ into sp_’||partition_name||’;’ xx into v_sql
    from (
    select rownum id, subpartition_name,partition_name
    from user_tab_subpartitions where table_name=’ACCOUNT_POSTING’ and partition_name=i.partition_name)
    start with id=1 connect by prior id=id-1
    group by partition_name;
   
    dbms_output.put_line(v_sql);

  end loop;
end;
/

Note that I learned the hard way that you can merge only 2 subpartitions at a time.

But still I will leave this note here, to illustrate the rows to csv conversion…

create table xml_repository
(file_name varchar2(30),
 xml clob);

Create a control file like this

load data
infile *
append into table xml_repository
(
file_name char(30),
xml lobfile (file_name) terminated by eof
)
begindata
file1.xml
file2.xml
file3.xml

Then prepare your xml files as listed above. This can be images or word document, etc (obviously with corresponding BLOB)

Then as usual

sqlldr userid/password@db control=loadxml.ctl

SELECT segment_name, segment_type, round(allocated_space/1024/1024,3) alloc_mb ,round(used_space/1024/1024,3) used_mb ,
round(reclaimable_space/1024/1024,3) reclaimable_mb, round(reclaimable_space/allocated_space*100,3) pct_wastage
FROM TABLE(dbms_space.asa_recommendations())
where segment_owner=’HILDA’
order by 6 desc

– find snapid where a particular sql occurs
select distinct snap_id from dba_hist_sqlbind where sql_id=’80bd6bb7n33tk’

– find bind variable values for the specified sql
select * from dba_hist_sqlbind
where snap_id = 20637 and trunc(last_captured) >= ‘17-NOV-2011′ and sql_id=’80bd6bb7n33tk’
order by sql_id, position

select snap_id, sample_time, session_id, session_serial#, sql_id, session_state, blocking_session, blocking_session_serial#, blocking_session_status, event from dba_hist_active_sess_history
where (session_id,session_serial#) in (select session_id,session_serial# from dba_hist_active_sess_history where sql_id=’0trpbjywfsdp6′ and snap_id between 20637 and 20639)
and snap_id between 20637 and 20639 order by snap_id, sample_time

4a2fkw1d4u01u — deletion profile that was blocked

select snap_id, sample_time, session_id, session_serial#, sql_id, session_state, blocking_session, blocking_session_serial#, blocking_session_status, event from dba_hist_active_sess_history
where (session_id,session_serial#) in (select session_id,session_serial# from dba_hist_active_sess_history where sql_id=’4a2fkw1d4u01u’ and snap_id between 20637 and 20639)
and snap_id between 20637 and 20639 order by snap_id, sample_time

select ‘USER’, sample_time, sql_id, session_state, blocking_session, blocking_session_serial#, blocking_session_status, event,
(select sql_text from v$sql where sql_id=a.sql_id and rownum=1) xxx from dba_hist_active_sess_history a
where session_id=762 and session_serial#=45940
and snap_id between 20637 and 20642 order by snap_id, sample_time

– blocked ETL Session
select ‘ETL’, sample_time, sql_id, session_state, blocking_session, blocking_session_serial#, blocking_session_status, event,
(select sql_text from DBA_HIST_SQLtext where sql_id=a.sql_id and rownum=1) xxx from dba_hist_active_sess_history a
where session_id=829 and session_serial#=31419
and snap_id between 20637 and 20642 order by snap_id, sample_time

–USER1@host2( SID=915 ) is blocking ETL_USER@host1( SID=829 )
– blocked by 915, which funnily enough only active 17 Nov from 17:44 till 22:15 - somehow this session did not release the lock at all until killed the following day
select ‘USER’, sample_time, sql_id, session_state, blocking_session, blocking_session_serial#, blocking_session_status, event,
(select sql_text from DBA_HIST_SQLtext where sql_id=a.sql_id and rownum=1) xxx from dba_hist_active_sess_history a
where session_id=915 and session_serial#=2165

select * from dba_hist_active_sess_history a
where session_id=915 and session_serial#=2165

select * from dba_tab_columns where column_name=’HOST_NAME’ and table_name like ‘DBA_HIST%’

select * from DBA_HIST_SQLbind where sql_id=’4a2fkw1d4u01u’ and snap_id between 20637 and 20642

——————————–
select a.*, round(index_leaf_estimate_if_rebuilt/current_leaf_blocks*100) percent, case when index_leaf_estimate_if_rebuilt/current_leaf_blocks < 0.5 then 'candidate for rebuild' end status
from
(
select table_name, index_name, current_leaf_blocks, round (100 / 90 * (ind_num_rows * (rowid_length + uniq_ind + 4) + sum((avg_col_len) * (tab_num_rows) ) ) / (8192 - 192) ) as index_leaf_estimate_if_rebuilt
from (
select tab.table_name, tab.num_rows tab_num_rows , decode(tab.partitioned,'YES',10,6) rowid_length , ind.index_name, ind.index_type, ind.num_rows ind_num_rows, ind.leaf_blocks as current_leaf_blocks,
decode(uniqueness,'UNIQUE',0,1) uniq_ind,ic.column_name as ind_column_name, tc.column_name , tc.avg_col_len
from dba_tables tab
join dba_indexes ind on ind.owner=tab.owner and ind.table_name=tab.table_name
join dba_ind_columns ic on ic.table_owner=tab.owner and ic.table_name=tab.table_name and ic.index_owner=tab.owner and ic.index_name=ind.index_name
join dba_tab_columns tc on tc.owner=tab.owner and tc.table_name=tab.table_name and tc.column_name=ic.column_name
where tab.owner='HILDA' and ind.leaf_blocks is not null and ind.leaf_blocks > 1000
) group by table_name, index_name, current_leaf_blocks, ind_num_rows, uniq_ind, rowid_length
) a where index_leaf_estimate_if_rebuilt/current_leaf_blocks < 0.5
order by index_leaf_estimate_if_rebuilt/current_leaf_blocks
——————————–

####################
# topinstance.sh
####################
#!/bin/ksh

DBS=`ps -eo args | grep pmon | grep -v grep | cut -c 10-`
PROCS=/tmp/_$$_procs.txt
ps -eoargs,pcpu > $PROCS

for DB in `echo $DBS`
do
echo `date` $DB `cat $PROCS | grep $DB | grep -v grep | sed ’s/.* *//’ | \
awk ‘BEGIN { totalcpu = 0.0 } { totalcpu += $0 } END { print totalcpu }’
`
done

rm -f $PROCS
#####################

This can then be combined with

####################
# pollstat.sh
####################
#!/bin/ksh
touch /home/users/hilda/hostname_usage.log

while [ true ]
do
/home/users/hilda/topinstance.sh >> /home/users/hilda/hostname_usage.log
sleep 30
done

===============

to grab a snapshot every 30 seconds.

This has proven to be useful when combined with AWR reports, so we can quickly determine whether a performance issue is self inflicted within an application, or because some other database is hogging the resource.

To realign the sequence, you can obviously drop/recreate/alter if you have the right privilege, however here is a tip of how to do it if you do not have the schema owner privilege.

declare
v_table number;
v_seq number;
v_diff number;
begin
select max(seq_id) into v_table from TABLE1;
select last_number into v_seq from all_sequences where sequence_owner=’OWNER01′ and sequence_name=’TABLE1_SEQ’ ;

select v_seq - v_table into v_diff from dual;

if v_diff < 0
then
select (v_diff * -1) + 2 into v_diff from dual;

for i in 1 .. v_diff
loop
select TABLE1_SEQ.nextval into v_table from dual;
end loop;

end if;

end;
/