I have a ssis project with 3 ssis packages, one is a parent package which calls the other 2 packages based on some condition. In the parent package I have a foreach loop container which will read multiple .csv files from a location and based on the file name one of the two child packages will be executed and the data is uploaded into the tables present in MS SQL Server 2008. Since multiple files are read, if any of the file generates an error in the the child packages, I have to log the details of error (like the filename, error message, row number etc) in a custom database table, delete all the records that got uploaded in the table and read the next file and the package should not stop for the files which are valid and doesn't generate any error when they are read.
Say if a file has 100 rows and there is a problem at row number 50, then we need to log the error details in a table, delete rows 1 to 49 which got uploaded in the database table and the package to start executing the next file.
How can I achieve this in SSIS?
You will have to set TransactionOption=*Required* on your foreach loop container and TransactionOption=*Supported* on the control flow items within it. This will allow for your transactions to be rolled back if any complications happen in your child packages. More information on 'TransactionOption' property can be found # http://msdn.microsoft.com/en-us/library/ms137690.aspx
Custom logging can be performed within the child packages by redirecting the error output of your destination to your preferred error destination. However, this redirection logging only occurs on insertion errors. So if you wish to catch errors that occur anywhere in your child package, you will have to set up an 'OnError' event handler or utilize the built-in error logging for SSIS (SSIS -> Logging..)
I suggest you try the creation of two dataflows in your loop container. The main idea here is to have a set of three tables to better and more easily handle the error situations. In the same flow you do the following:
1st dataflow:
Should read .csv file and load data to a temp table. If the file is processed with errors you simply truncate the temp table. In addition, you should also configure the flat file source output to redirect the errors to an error log table.
2nd dataflow:
On the other hand, in case of processing error-free, you need to transfer the rows from temp into the destination table. So, here, the OLEDB datasource is "temp table" and the OLEDB destination is "final table".
Don´t forget to truncate the temp table in both cases, as the next file will need an empty table.
Let's break this down a bit.
I assume that you have a data flow that processes an individual file at a time. The data flow would read the input file via a source connection, transform it and then load the data into the destination. You would basically need to implement the Error Handler flow in your transformations by choosing "Redirect Row". Details on the Error Flow are available here: https://learn.microsoft.com/en-us/sql/integration-services/data-flow/error-handling-in-data.
If you need to skip an entire file due to a bad format, you will need to implement a Precedence Constraint for failure on the file system task.
My suggestion would be to get a copy of the exam preparation book for exam 70-463 - it has great practice examples on exactly the kind of scenarios that you have run into.
We do something similar with Excel files
We have an ErrorsFound variable which is reset each time a new file is read within the for each loop.
A script component validates each row of the data and sets the ErrorsFound variable to true if an error is found, and builds up a string containing any error details.
Then - based on the ErrorsFound variable - either the data is imported or the error is recorded in a log table.
It gets a bit more tricky when the Excel files are filled in badly enough for the process not to be able to read them at all - for example when text is entered in a date, number or currency field. In this case we use the OnError Event Handler of the Data flow task to record an error in the log but won't know which row(s) caused the problem
Related
My goal is to write error records from both source and destination into one file.
Currently, I'm encountering
Warning: The process cannot access the file because it is being used by another process.
PS: As I explore this feature, I can write these error records into two separate files.
Instead of defining two flat file destinations, define just one and union the error outputs to it.
I have a SSIS Package that I have a For Each Loop which imports multiple txt files into a SQL Server table. That runs fine.
What I am trying to accomplish is to store the distinct filename and date it was imported into a separate table. I created a separate For Each Loop for this and then archive the txt file after it's complete with a File System Task.
The issue I am having is I put an event handler to invoke a SQL Task and Send Email task if there is a warning (I was hoping for a warning only if there were no files in the directory where the package is importing from).
However, I found a warning that a column in the Data Flow task was not being used and should be removed if not needed. But the Data Flow task requires at least one field for me to put a Derived Column task
Derived Column Field1: pulls the #User: CurrentFile from the ForEachLoop Container.
Field2 pulls the current date.
Is there a way to perform this without the warning?
It sounds like you're over-complicating thing.
You have a ForEach loop and you're therefore assigning a value into some Variable to contain the file name, #User::CurrentFile. You can get the date it was loaded through either a call to GETDATE() or reference the system scoped variable, StartTime #[System::StarTime]
The most straight forward option would be to add an Execute SQL Task wired up to the OnSuccess Precedent Constraint from your Data Flow Task. The Execute SQL Task will then have a statement like INSERT INTO dbo.MyLog(FileName, InsertDate) SELECT ?, ?, assuming OLE DB Connection Manger, and then you map in your two variables.
Easy, clean, no warnings fired about unused columns in your data flow.
What I think you have is something like this, based on
I created a separate For Each Loop for this
I have SSIS packages to extract fact tables into the staging tables. I have a control table which contains the last extract date for each table. So, the package extract rows where > control table date. The problem I have is I want to redirect rows with error to an error file in the data flow task of the package. If I do that the package will not fail (so I can't rollback) and some rows might actually go through which if I coninue with the process will ultimately get to my fact table. Now, next time when I run the package if I had updated the control table, I will miss the rows which had erros. If I had not updated the control table with the date, I will re-extract the rows which went through. What is the best practice for this?
How about adding a Row Count Transformation onto the error branch? It sounds like you are using the transaction option in SSIS so put the Data Flow in a sequence container and post Data Flow, evaluate the value of your row count variable. If it's greater than zero, rollback/abort processing.
I'm new to SSIS. I m trying to load data from excel to sql server. Each sheet in excel should go into separate table in sql server so I created a sequence container and inside that I have placed data flow tasks. Each data flow task contains logic to move data from one excel sheet to table in sql server. So problem is if one data flow task fails then I was expecting whole package will fail and rollback the changes done by other data flow tasks. But its not doing that. Other data flow tasks are getting committed into sql server.
I have set the property "Failpackageonfailure" to true for sequence container and individual data flow tasks. Also set the transaction as "Required" but no change in results.
Please help!
Thanks in advance
Sai
You could have also set Required to a newly created sequence container with the control flow items inside of it (in case your package does other tasks that do not behave the same way).
I am running a SSIS package using SQL Server 2008 Job. The package crash at some point while running. I have created my own mechanism to grab the error and record it in a table. So I can see that there is an error with an specific task, but could not find what the error is.
When I run the same package from BIDS, it works perfect. no error.
What I want to do is, I need to write that error string to my own table which shown in the "Execution Result" tab.
So the question is which system variable holds the error string in SSIS.
The error is stored in the ErrorDescription system variable. See Handling Errors in the Data Flow for an example of how to get the error description.
Also, if you want to capture error information into a table, SSIS supports logging to a table using the SQL Server Log Provider. You can also customize the logging.
Too easy.
Left-Click (highlight) on the object you want to capture the error event (Script, or Data Flow, etc.)
Click on 'Event Handlers' - screen should open with Executable = object you clicked and Event Handler = OnError
Click URL (click here to create....)
Drag Execute SQL object from SSIS Toolbox
Configure to the database/table you want to house the error message
Write INSERT INTO DB.Schema.Table(DBName, SchemaName, TableName,ErrorMessage,DateAdded)
Write VALUES (?,?,?,'I am smart',getdate())
Click Parameters and select the USER::Variables for the ?'s + my comment.
Since this is ran at the database server it will pass in the ?'s. My SAC is already at the database as a value but you will have selected System::ErrorDescription as parameter 3. Remember, this array is 0 based. DO NOT TRY TO NAME THE PARAMETERS. Instead, number them 0 to ~? The datatypes are based on what you have going in; mine are all VARCHAR so... :)
This is a much better solution than just logging whatever the server allows you to.
I can also add a counter variable and adjust it wherever I like; then pass it to the event OnError. This will allow me to pinpoint exactly where the last successful object completed; works best in scripting objects but also available in other areas.
I'm using this so I can process thousands of cycles without actually failing the package. If a table doesn't exist or a column doesn't exist I simply log it for further review later. Oh yeah, I'm cycling through hundreds of databases capturing their architecture and maximum column size used; not to be confused with maximum column size.
Example: TelephoneNumber comes from a source column of char(500) (definitely bad programming but...you can't change everything so..). I capture the max len of that column and adjust the destination column to accommodate that size +/- a certain percentage.
If a table doesn't exist or a column doesn't exist anymore I log the error and keep churning. At the end, I can evaluate those entries and see if I can actually remove them from my warehouse. This happens more in the TEST and STAGE environments than in PROD. However, when a change goes through to PROD I most definitely will identify it as it's coming in to the warehouse.
Everything is configured, this includes dynamic MERGE/JOINs, INSERT, SELECT, ELEMENTS, SIZES, USAGESIZE, IDENTITY, SOURCEORDER, etc. with conversions of data to destination datatypes.
ALL that because the systemic version of logging will not provide you with the granularity you might need for this type of operation. This OnError Event Handler can if setup properly.
Check this out! He has explained with a Step by step process on how to configure SSIS logging which has the error message parameter.