asked on

SSIS ETL package to update/insert staging source to destination ?

Hello,

Could someone please help me here, new to SSIS and ETL stuff, I have raw data being staged to a SQL table, which I then want to move to our production table and update/insert the record set based on a unique key.

The staging table has a row "SEQ" that identifies the record set, based on the SEQ I want to compare the destination table and if the SEQ exists, update the whole record or insert new otherwise.

I am using SSIS to acomplish this as per the image attached.

1. Connect OLE DB Source
2. Lookup transformation Editor, and configure error output to ignore failure.
3. Conditional Split
      - Insert Record Condition: ISNULL(Dest_SEQ)
      - Update Record Condition: (Dest_TYP != TYP) || (Dest_SALEDATE != SALEDATE) || (Dest_INVOICED != INVOICED) || (Dest_WEEKDATE != WEEKDATE) || (Dest_EWK != EWK) || (Dest_ZWK != ZWK) || (Dest_IY != IY) || (Dest_WK != WK) || (Dest_ROMO != ROMO) || (Dest_ROYR != ROYR)

4. New Reords go to OLE DB Destination
5. Updated records go to OLE DB Command as such:

UPDATE dbo.sales
SET
TYP = ?,
SALEDATE = ?,
INVOICED = ?,
WEEKDATE = ?,
EWK = ?,
ZWK = ?,
IY = ?,
WK = ?,
ROMO = ?,
ROYR = ?
WHERE SEQ = ?

The staging table has about 100 columns with data, so I am not sure how appropriate it is to define a Update Record condition for each one of those, perhaps my update record will be very expensive as far as resource when we are talking about 100,000 updates..

Is there a way I can change this so that in a conditional split (or using something else) to check if staging's SEQ exists in destination, if it does, update the whole record.

The above works, but ideally should I be creating a primary key on my SEQ table, and then doing an update/insert based on if the primary key exists or not, how could I accomplish this?

Right now, if I set the SEQ as a primary key on my "sales" destination table, and re-run my SSIS I get an error:

An OLE DB record is available. Source: "Microsoft SQL Native Client" Hresult: 0x80004005 Description: "Violation of PRIMARY KEY constraint 'PK_sales_1'. Cannot insert duplicate key in object 'dbo.sales'.".

But the table is empty, why can't it insert the SEQ key?

Thanks for all your help.

The guide I followed was: http://vsteamsystemcentral.com/cs21/blogs/applied_business_intelligence/archive/2007/05/21/ssis-design-pattern-incremental-loads.aspx

Staging table layout:

USE [Sales]
CREATE TABLE [dbo].[stg_sales](
	[SEQ] [int] NOT NULL,
	[TYP] [varchar](1) NULL,
	[SALEDATE] [varchar](50) NULL,
	[INVOICED] [varchar](50) NULL,
	[WEEKDATE] [varchar](50) NULL,
	[EWK] [varchar](50) NULL,
	[ZWK] [varchar](50) NULL,
	[IY] [varchar](50) NULL,
	[WK] [varchar](50) NULL,
	[ROMO] [varchar](50) NULL,
	[ROYR] [varchar](50) NULL,
	[YRTD] [varchar](50) NULL,
GO

Destination Table Layout:

USE [Sales]
CREATE TABLE [dbo].[sales](
	[SEQ] [int] NOT NULL,
	[TYP] [varchar](1) NULL,
	[SALEDATE] [varchar](50) NULL,
	[INVOICED] [varchar](50) NULL,
	[WEEKDATE] [varchar](50) NULL,
	[EWK] [varchar](50) NULL,
	[ZWK] [varchar](50) NULL,
	[IY] [varchar](50) NULL,
	[WK] [varchar](50) NULL,
	[ROMO] [varchar](50) NULL,
	[ROYR] [varchar](50) NULL,
	[YRTD] [varchar](50) NULL,
GO

Open in new window

SSIS.JPG
lookup-transorm.jpg
conditional-split.jpg

ASKER CERTIFIED SOLUTION

valkyrie_nc

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial

Reza Rad

I have a question:
do you want to look for SEQ in staging table, and if you find update it , if don't find insert?

if yest, why you don't use UNMATCH result of Lookup Transformation.
match result of lookup means rows which currently exists and need to be updated so redirect them to OLE DB Command.
un match result of lookup means rows which not exists currently and you should insert them, so redirect them to Destination.

does it make sense to you ?
please let me know if you mean anything else.

@valkyrie_nc:
SCD ( SLowly Changing Dimenstion ) hasn't good performance, and this will be better to use other ways if possible. Of course the logic of question isn't very clear to me yet, but I think there are better ways than SCD.

mirde

ASKER

The SCD seems to work for me, as we will only be importing 10,000 - 30,000 new/updated records per day, this is a very "small" load for SQL Server/SCD.