DB2 - Querying two databases in SQL PL - database

Question

I need to compare data across two different DB2 database instances. We're not permitted to set up federation. I've found references saying how to specify data loads from remote databases, and also references on how to specify a database connection including database name, username etc. Ideally I would be able to execute a query against one database, then compare that to the second database either one-by-one (using SQL PL loops etc.), or as a single large join. I've gotten to the point where the SQL PL script can connect to each in turn (and it prompts me for the password to both), but it only recognizes the second one when I attempt to query the table.
What we've tried:
Adding two different CONNECT statements at the beginning.
Declaring a cursor and specifying the database name (this seems to only work when doing loads from one database to another, which we're trying to avoid).
set serveroutput on#
set sqlcompat DB2#
connect to first user myname#
connect to second user myname#
-- run command: db2 -td# -vf test3.sql
begin
declare loop_counter int;
call dbms_output.enable(100000);
set loop_counter = 0;
FIRSTLOOP: for o as ord1 cursor for
select field1, field2 from first.firstschema.firsttable fetch first 10 rows only with ur
do
set loop_counter = loop_counter + 1;
call dbms_output.put_line('Field: '||field1||', other '||field2);
end for;
call dbms_output.put_line('End first program: ');
SECONDLOOP: for p as ord2 cursor for
select field1, field2 from second.secondschema.secondtable fetch first 10 rows only with ur
do
set loop_counter = loop_counter + 1;
call dbms_output.put_line('Field: '||field1||', other '||field2);
end for;
call dbms_output.put_line('After second call');
end#
Ideally, each of the two cursor loops would print 10 rows. In reality, whichever CONNECT was done second is the one that works. For example, if I have the connect to SECOND followed by the connect to FIRST, the first loop works and the second says "..... is an undefined name". If I do the connect to FIRST then the connect to SECOND, the first loop throws the error and I get no output.

Answer 1

SQL PL can connnect only to one database at a time - that is the design.
In your script example, the second connect will close any current connection first.
Federation lets you access remote tables as if they were local.
If you are prevented from using federation , your options include these:
materialising the remote table locally and copying the data
(this can be done via load from remote cursor).
You can then use SQL to compare rows, as both tables are then in the same database.
This is only feasible if you have sufficient capacity to fit both tables in same database, although compression will help here.
not using SQL but instead using another tool
For example: depending on data volumes , and data-types, you could export source/target tables
to flat files and compare the files (diff etc). You could also export to pipe and use in memory comparisons.
Or you could use python or perl or any scripting language and do the comparison in memory in chunks (in all cases
each thread can only connect to a single database at one time).
use third party tools for data comparison.
if you use embedded-SQL , type-2 connect offers another possibility.

Answer 2

On Db2 for IBM i, federation is only available via Db2 LUW box...
However, the following works in Db2 for IBM i...
create or replace function myschema.myudtf ()
returns table (SERVER VARCHAR(18)
, as_of timestamp
, ORDINAL_POSITION INTEGER
, JOB_NAME VARCHAR(28)
, SUBSYSTEM VARCHAR(10)
, AUTHORIZATION_NAME VARCHAR(10)
, JOB_TYPE VARCHAR(3)
)
modifies SQL data
external action
not deterministic
language SQL
specific CHKAWSJOBS
begin
declare insertStmt varchar(1500);
declare global temporary table
GLOBAL_TEMP_MY_JOBS (
SERVER VARCHAR(18)
, as_of timestamp
, ORDINAL_POSITION INTEGER
, JOB_NAME VARCHAR(28)
, SUBSYSTEM VARCHAR(10)
, AUTHORIZATION_NAME VARCHAR(10)
, JOB_TYPE VARCHAR(3)
) with replace;
for systemLoop as systemsCursor cursor for
select * from table( values ('mysys1'),('mysys2'),('mysys3'))
as systems (server_Name)
do
set insertStmt =
' insert into GLOBAL_TEMP_MY_JOBS
select
current_server as server, current_timestamp as as_of
, ordinal_position, job_name, subsystem, authorization_name, job_type
from table(QSYS2.ACTIVE_JOB_INFO(
SUBSYSTEM_LIST_FILTER => ''MYSBS'')) X
where exists (select 1 from ' concat server_name concat '.sysibm.sysdummy1)';
execute immediate InsertStmt;
end for;
return select * from GLOBAL_TEMP_MY_JOBS;
end;
The example above is more complex than your use case, I'm pulling data from a UDTF on the remote system, the trick is the use of a 3 part name in the where clause, which forces the DB to run the entire select statement on the remote machine; with the insert being into the table on the local machine.
You should be able to build a dynamic insert that's just
set insertStmt = 'insert into lcltable
select field1, field2
from ' concat server_name concat table_name
concat ' fetch first 10 rows only with ur';
Don't know for sure that this will work on Db2 LUW, but there's a good chance.