Error handling

Question

Error handling

brian.geregach

SSC-Addicted

Points: 406
More actions
April 16, 2013 at 7:58 am

#276502

First off, my appologies, I am a newbie to programming and SQL.
Setup:
I have an SSIS package that inserts rows to a database. On occasion I get a constraint error on one of my reference tables. So far it is because of missing data.
What I want to do:
I want to redirect the row, grab the data and insert it into the reference table then reprocess the row.
Questions:
A) is this possible? I have looked at doing a lookup task prior to the row being inserted to the database. I have also looked at using a sql task in my error handling, but having issues with the code (see B).
B) I need help with the code. How do I pull the information from that single row then insert it into the database? Do I use a variable? Is this a select statement?
Any help would be greatly appreciated.
Thank you
Brian

Viewing 5 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic. Login to reply

Mansfield SSCrazy Points: 2396 More actions · Answer 1

I would set up the SSIS package to insert into a staging table first. Then use a SQL task to check for references in the staging table that does not yet exist, and insert the references. Use something like a left outer join check:

insert into reference_table (reference_id, reference_description)

select s.reference_id, 'some description'

from staging_table s

left join reference_table r on r.reference_id = s.reference_id

where r.reference_id is null

After the references are resolved, copy from the staging table to the actual table being loaded.

Phil Parkin SSC Guru Points: 246977 More actions · Answer 2

brian.geregach (4/16/2013)
First off, my appologies, I am a newbie to programming and SQL.
Setup:
I have an SSIS package that inserts rows to a database. On occasion I get a constraint error on one of my reference tables. So far it is because of missing data.
What I want to do:
I want to redirect the row, grab the data and insert it into the reference table then reprocess the row.
Questions:
A) is this possible? I have looked at doing a lookup task prior to the row being inserted to the database. I have also looked at using a sql task in my error handling, but having issues with the code (see B).
B) I need help with the code. How do I pull the information from that single row then insert it into the database? Do I use a variable? Is this a select statement?
Any help would be greatly appreciated.
Thank you
Brian

There is more than one way to solve this. Here are three:

1) (Pure SSIS)

Use a lookup task to check for existence. When matched then continue as currently. When not matched:

a) Insert a suitable row into the reference table using a OLEDB command

b) Insert to child table as usual.

2) (Hybrid)

Use a lookup as above. When matched, all OK. When not matched

a) Direct to a staging table

b) At the end of the dataflow, run a stored proc to process and insert the rows in the reference and staging tables as required.

3) (Use SSIS as little as possible)

Use the solution suggested by Mansfield.

I would expect (2) to be the fastest, (1) is the most elegant and (3) is probably easiest to set up, maintain and troubleshoot - for SQL folks anyway. In my opinion, of course.

One problem with (1) is that you cannot use a cached lookup - so performance is likely to be average at best.

The absence of evidence is not evidence of absence.
Martin Rees

You can lead a horse to water, but a pencil must be lead.
Stan Laurel

brian.geregach SSC-Addicted Points: 406 More actions · Answer 3

My thanks to the two responders. Both suggestions look very helpful.

Brian

Sergiy SSC Guru Points: 110208 More actions · Answer 4

Mansfield (4/28/2013)
I would set up the SSIS package to insert into a staging table first. Then use a SQL task to check for references in the staging table that does not yet exist, and insert the references. Use something like a left outer join check:
insert into reference_table (reference_id, reference_description)
select s.reference_id, 'some description'
from staging_table s
left join reference_table r on r.reference_id = s.reference_id
where r.reference_id is null
After the references are resolved, copy from the staging table to the actual table being loaded.

From my experience WHERE NOT EXISTS always perfoms better, at least not worse, than LEFT JOIN with IS NULL check:

insert into reference_table (reference_id, reference_description)

select s.reference_id, 'some description'

from staging_table s

WHERE NOT EXISTS (select * from reference_table r

where r.reference_id = s.reference_id)

_____________
Code for TallyGenerator