Question

I’m trying to create a script to migrate data from one DB to another. One thing I’m not currently able to do is set the nextval of a sequence to the nextval of a sequence in another DB.

I got the difference in values from user_sequences and generated the following dynamic SQL statements:

execute immediate 'alter sequence myseq increment by 100';
execute immediate 'select myseq.nextval from dual';
execute immediate 'alter sequence myseq increment by 1';

commit;

But nothing happens. What am I missing? If I run the same statements outside the procedure, they work fine:

alter sequence myseq increment by 100;
select myseq.nextval from dual;
alter sequence myseq increment by 1;

commit;

EDIT: Apologies to all for not being clear. I’m actually altering the sequence in the same DB. I’m only getting the value to be set from a remote DB. Perhaps it was unnecessary to mention the remote DB as it doesn’t affect things. I only mentioned it to explain what my goals were.

Step 1. I get the nextval of the sequence from a remote DB.

select (select last_number
        from dba_sequences@remoteDB
        where upper(sequence_name) = upper(v_sequence_name)) - (select last_number
                                                                from user_sequences
                                                                where upper(sequence_name) = upper(v_sequence_name)) increment_by
from dual;    

Step 2. I generate dynamic SQL statements with this value:

execute immediate 'alter sequence myseq increment by 100';
execute immediate 'select myseq.nextval from dual';
execute immediate 'alter sequence myseq increment by 1';

commit;

No error was raised, but nothing happened. When I wrote the SQL statements with DBMS_OUTPUT.PUT_LINE and ran them outside they worked.

Answer 1

Here is some code which dynamically sets a sequence to a new (higher) value. I have written this so it will work for any sequence in your schema.

create or replace procedure resync_seq
    (p_seq_name in user_sequences.sequence_name%type)
is
    local_val pls_integer;
    remote_val pls_integer;
    diff pls_integer;
begin
    execute immediate 'select '|| p_seq_name ||'.nextval from dual'
           into local_val;
    select last_number into remote_val
    from user_sequences@remote_db
    where sequence_name = p_seq_name ;
    diff := remote_val - local_val;

    if diff > 0
    then
        execute immediate 'alter sequence  '|| p_seq_name ||' increment by ' ||to_char(diff);
        execute immediate 'select '|| p_seq_name ||'.nextval from dual'
           into local_val;
        execute immediate 'alter sequence  '|| p_seq_name ||' increment by 1';
    end if;

end;

The procedure doesn’t need a COMMIT because DDL statements issue an implicit commit (two in fact).

You can execute it and see the synced value like this (in SQL*PLus):

exec resync_seq('MYSEQ')
select myseq.currval
from dual

Incidentally, the only way to reset a sequence (to its original starting value or a different lower value) is dropping and re-creating the sequence.

---

In 18c Oracle added a RESTART capability to ALTER SEQUENCE. The straightforward option …

alter sequence myseq restart;

…resets the sequence to the value specified by the START WITH clause in the original CREATE SEQUENCE statement. The other option allows us to specify a new starting point:

alter sequence myseq restart start with 23000;

Excitingly this new starting point can be ahead or behind the current value (within the usual bounds of a sequence).

The one snag is that this new capability is undocumented (only for Oracle’s internal usage) and so we’re not supposed use it. Still true in 20c. The only approved mechanism for changing a sequence’s value is what I outlined above.