Transaction is rolled back after commit? - sql-server

I'm experiencing some problems that look a LOT like a transaction in a stored procedure has been rolled back, even though I'm fairly certain that it was committed, since the output variable isn't set until after the commit, and the user gets the value of the output variable (I know, because they print it out and I also set up a log table where i input the value of the output variable).
In theory someone COULD manually delete and update the data such that it would look like a rollback, but it is extremely unlikely.
So, I'm hoping someone can spot some kind of structural mistake in my stored procedure. Meet BOB:
CREATE procedure [dbo].[BOB] (#output_id int OUTPUT, #output_msg varchar(255) OUTPUT)
as
BEGIN
SET NOCOUNT ON
DECLARE #id int
DECLARE #record_id int
SET #output_id = 1
-- some preliminary if-statements that doesn't alter any data, but might do a RETURN
SET XACT_ABORT ON
BEGIN TRANSACTION
BEGIN TRY
--insert into table A
SET #id = SCOPE_IDENTITY()
--update table B
DECLARE csr cursor local FOR
SELECT [some stuff] and record_id
FROM temp_table_that_is_not_actually_a_temporary_table
open csr
fetch next from csr into [some variables], #record_id
while ##fetch_status=0
begin
--check type of item + if valid
IF (something)
BEGIN
SET SOME VARIABLE
END
ELSE
BEGIN
ROLLBACK TRANSACTION
SET #output_msg = 'item does not exist'
SET #output_id = 0
RETURN
END
--update table C
--update table D
--insert into table E
--execute some other stored procedure (without transactions)
if (something)
begin
--insert into table F
--update table C again
end
DELETE FROM temp_table_that_is_not_actually_a_temporary_table WHERE record_id=#record_id
fetch next from csr into [some variables], #record_id
end
close csr
deallocate csr
COMMIT TRANSACTION
SET #output_msg = 'ok'
SET #output_id = #id
END TRY
BEGIN CATCH
ROLLBACK TRANSACTION
SET #output_msg = 'transaction failed !'
SET #output_id = 0
INSERT INTO errors (record_time, sp_name, sp_msg, error_msg)
VALUES (getdate(), 'BOB', #output_msg, error_message())
END CATCH
RETURN
END
I know, my user gets an #output_id that is the SCOPE_IDENTITY() and he also gets an #output_msg that says 'ok'. Is there ANY way he can get those outputs without the transaction getting committed?
Thank you.

You know the problem is that transaction dose NOT support rollback on variables because there is no data change inside database. Either commit or rollback of the transactions ONLY make difference on those database objects (tables, temp table, etc.), NOT THE VARIABLES (including table variables).
--EDIT
declare #v1 int = 0, #v2 int = 0, #v3 int = 0
set #v2 = 1
begin tran
set #v1 = 1
commit tran
begin tran
set #v3 = 1
rollback tran
select #v1 as v1, #v2 as v2, #v3 as v3
RESULT is as follows

Personally I never used transactions in stored procedures, especially when they are used simultaniously by many people. I seriously avoid cursors as well.
I think I would go with passing the involved rows of temp_table_that_is_not_actually_a_temporary_table into a real temp table and then go with an if statement for all rows together. That's so simple in tsql:
select (data) into #temp from (normal_table) where (conditions).
What's the point of checking each row, doing the job and then rollback the whole thing if say the last row doesn't meet the condition? Do the check for all of them at once, do the job for all of them at once. That's what sql is all about.

Related

SQL Server - Changes not reflecting everytime until manually doing CHECKPOINT

I am having an issue in SQL Server procedure.
I have two new stored procedures, with the PROC_Main proc performing a bunch of inserts and updates before it calls the PROC_child to pull the updated records back out.
--Child PROC
CREATE PROCEDURE dbo.Proc_Child
#Id int
AS
BEGIN
SELECT * FROM dbo.Employee WHERE Id = #Id AND Status=1
END
--Parent Proc
CREATE PROCEDURE dbo.Proc_Main
#Id int ,#Status varchar(100),#Date datetime
AS
BEGIN
BEGIN TRY
BEGIN TRAN
IF NOT EXISTS (SELECT Id FROM dbo.Employee WHERE Id = #Id)
BEGIN
UPDATE dbo.Employee
SET Status = 3,
Date = getdate()
WHERE Status <> 3
AND Id = #Id
INSERT INTO dbo.Employee (ID,Status,Date)
VALUES (#ID,#Status,#Date)
END
COMMIT
--CHECKPOINT;
EXEC dbo.Proc_Child #Id = #Id
END TRY
BEGIN CATCH
IF ##TRANCOUNT > 0
ROLLBACK TRAN
DECLARE #Message VARCHAR(1000) = ERROR_MESSAGE()
DECLARE #Severity INT = ERROR_SEVERITY()
DECLARE #State INT = ERROR_STATE()
RAISERROR(#Message, #Severity, #State)
END CATCH
END
--Procedure call
EXEC Proc_Main #ID=1,#Status=1,#Date='2019-01-01'
I am facing the issue that Proc_Main is not returning the records from PROC_Child every time.
When I am manually doing checkpoint before Proc_Child is called then only it is returning records.
Nothing to do with checkpoint. Based on your code, if you call main proc with Status != 1, your child proc will not return it. Also, why are you doing update if you know that record does not exist? Finally, in the multi-threaded environment this may blow up, you need to lock the id when you checking for the existence.

Select a column from a row and update the column to another value

I am trying to write a stored procedure that reads a column in a particular row of a table, then updates that column with a new value. The orig. is returned.
I want it to lock the row from others till I am done. What is the process?
I have something like
CREATE PROCEDURE [dbo].[aptc_Prt_NextDocumentNumberGet]
(#_iFormatConfigID INT, #_oNextDocumentNumber FLOAT OUTPUT)
AS
BEGIN
DECLARE #FrameworkConfig XML
SET #_oNextDocumentNumber = - 1
DECLARE #NewNextDocumentID FLOAT
SELECT
#_oNextDocumentNumber = FrameworkConfig.value('(/Parameters/Parameter[#Name="NextDocNo.NextDocumentNumber"])[1]', 'float')
FROM
[ttcPrtFormatConfig] WITH (ROWLOCK)
WHERE
FormatConfigID = #_iFormatConfigID
-- Select the Next Doc num out of the xml field
-- increment appropriate control and set output
IF #_iFormatConfigID IS NOT NULL
BEGIN
-- set what will be the "next" doc number after we add this current txn
IF (ABS(#_oNextDocumentNumber - 99999999999999999) < 0.0001)
BEGIN
SELECT #NewNextDocumentID = 1
END
ELSE
BEGIN
SELECT #NewNextDocumentID = #_oNextDocumentNumber + 1
END
UPDATE [ttcPrtFormatConfig]
WITH (ROWLOCK)
SET FrameworkConfig.modify('
replace value of
(/Parameters/Parameter[#Name="NextDocNo.NextDocumentNumber"]/text())[1]
with sql:variable("#NewNextDocumentID")')
WHERE FormatConfigID = #_iFormatConfigID
END
END
This should get you close to what you want.
DECLARE #MyValue INT
--You need a transaction so that the scope of your lock is well defined
BEGIN TRANSACTION
BEGIN TRY
--Get the value you are interested in, This select will lock the row so other people will not even be able to read it until you are finished!!!!!
SELECT #MyValue = MyValue
FROM MyTable WITH (UPDLOCK HOLDLOCK)
WHERE MyValue = SomeValue
--Do your checks and updates. You can take as long as you like as you are the only person who can do a read or update of this data.
IF
BEGIN
UPDATE MyTable
END
--Make sure you commit or rollback! this will release the lock
END TRY
BEGIN CATCH
--Oh no bad stuff! give up and put it back to how it was
PRINT ERROR_MESSAGE() + N' Your message here'
--Check there is a transaction that we can rollback
IF ##TRANCOUNT > 0
BEGIN
ROLLBACK;
END
--You may want to return some error state and not throw!
THROW;
--RETURN -1 --(for example)
END CATCH;
--yay it all worked and your lock will be released
COMMIT
--Do what you like with the old value
RETURN #MyValue

Return Resultset from SQL Server to VB.NET application

I need to return a resultset consisting of database errors from a SQL Server stored procedure's CATCH clause but I'm stuck with it. Do I need to use cursors to return resultset and if so, then what is the type declaration for the OUTPUT parameter in my .NET application? I tried Object and Variant but did not work.
I also tried the simple way just using a SELECT statement to return and it works with one stored procedure but not with another as thus in my CATCH clause:
while (#I <= #count)
begin
BEGIN TRY
-- delete all previous rows inserted in #customerRow for previous counts #I
delete from #customerRow
-- this is inserting the current row that we want to save in database
insert into #customerRow
SELECT
[id],[firstname], [lastname], [street], [city],
[phone],[mobile],[fax], [email], [companyName],
[licence],[brn], [vat], [companyStreet], [companyCity], [status]
FROM
(SELECT
ROW_NUMBER() OVER (ORDER BY id ASC) AS rownumber,
[id], [firstname], [lastname], [street], [city],
[phone], [mobile], [fax], [email], [companyName],
[licence], [brn], [vat], [companyStreet], [companyCity], [status]
FROM
#registerDetails) AS foo
WHERE
rownumber = #I
-- this stored procedure handles the saving of the current customer row just defined above
-- if there is any error from that sproc, it will jump to CATCH block
--save the error message in the temp table and continue
--with the next customer row in the while loop.
exec dbo.sp_SaveCustomer #customerRow
END TRY
BEGIN CATCH
IF ##TranCount = 0
-- Transaction started in procedure.
-- Roll back complete transaction.
ROLLBACK TRANSACTION;
if XACT_STATE()= -1 rollback transaction
DECLARE #ErrorMessage NVARCHAR(4000);
DECLARE #ErrorSeverity INT;
DECLARE #ErrorState INT;
SELECT #ErrorMessage = ERROR_MESSAGE() + ' ' + (select firstname from #registerDetails where id=#I)
SELECT #ErrorSeverity = ERROR_SEVERITY();
SELECT #ErrorState = ERROR_STATE()
INSERT INTO #registrationResults (error,id)
SELECT #ErrorMessage, #I
END CATCH
set #I = #I +1
end
COMMIT TRANSACTION registerTran
select * from #registrationResults
The above works with one stored procedure when I call it in my vb.net code as :
ta.Fill(registrationErrors, clientDetailsDT)
where registrationErrors and clientDetailsDT are strongly typed data tables.
This one does not :
begin catch
IF ##TranCount > 0 or XACT_STATE()= -1 ROLLBACK TRANSACTION;
DECLARE #ErrorMessage NVARCHAR(4000);
DECLARE #ErrorSeverity INT;
DECLARE #ErrorState INT;
DECLARE #ErrorLine INT;
SELECT #ErrorMessage = ERROR_MESSAGE();
SELECT #ErrorSeverity = ERROR_SEVERITY();
SELECT #ErrorState = ERROR_STATE();
SELECT #ErrorLine = ERROR_Line();
****ERROR -- THIS SELECT WAS MESSING ALL UP as it was this select that was being returned to the .NET and not the select of the desired #temp table after, hence returning 0 resultset as this select was EMPTY. !!
select status_indicator from InsertInvoiceTriggerData where session_GUID = guid**
delete from InsertInvoiceTriggerData where session_GUID = #guid**
INSERT INTO #registrationResults (error,id)
SELECT #ErrorMessage, NULL
select * from #registrationResults
end catch
Any suggestions how to return resultsets?
I haven't seen your database code, but in my experience the very first error caught by catch means that the entire transaction has to be rolled back. Apart from other things, it also implies that I never have more than 1 error to return in any given situation.
As such, I use 2 scalar output parameters in my stored procedures, that is:
#Error int = null output,
#Message nvarchar(2048) = null output
And I can retrieve them just like any other output variables.
UPD: Even after you have added some code, I still fail to understand what is your problem, exactly. However, I see several problems with your code, so I'll point them out and chances are, one of them will solve the problem. I am commenting only the first snippet, since the last one is too incomplete.
You should have been started the outermost transaction somewhere before the loop. If not, the code will fail.
If I guessed correctly, you implemented all savepoint logic inside the dbo.sp_SaveCustomer stored proc. If not, the whole discussion is pointless, since there are no save tran or rollback #savepoint statements in the code you have shown.
The first catch statement - IF ##TranCount = 0 ROLLBACK TRANSACTION is all wrong. If the condition is successful, it will result in error trying to rollback nonexistent transaction. Should not be here if you rely on savepoints.
The next after it should result in unconditional break:
if XACT_STATE()= -1 begin
rollback transaction;
break;
end;
The rest of your catch code can be replaced with this:
INSERT INTO #registrationResults (error, id)
SELECT error_message() + ' ' + firstname, id
from #registerDetails where id=#I;
Also, never use temp tables for this purpose, because rollback will affect them as well. Always use table variables for this, they are non-transactional (just like any other variable).
The commit should be conditional, because you may end up at this point with no transaction to commit:
if ##trancount > 0 commit tran;
There is no point in specifying savepoint name in the commit statement, it only leads to confusion (though isn't considered an error). Also, there should not any savepoint in this module (unless you have defined it before the loop).
I suspect that's just the tip of the iceberg, since I have no idea what actually happens inside the dbo.SaveCustomer stored procedure.
UPD2: Here is a sample of my VB.NET code which I use to receive recordsets from stored procedures:
Private Function SearchObjectsBase( _
SearchMode As penum_SEARCH_MODE, SearchCriteria As String
) As DataSet
Dim Cmd As DbCommand, Pr As DbParameter, dda As DbDataAdapter
' Initialise returning dataset object
SearchObjectsBase = New DataSet()
Cmd = MyBase.CreateCommand(String.Format("dbo.{0}", SearchMode))
With Cmd
' Parameter definitions
Pr = .CreateParameter()
With Pr
.ParameterName = "#SearchCriteria"
.DbType = DbType.Xml
.Value = SearchCriteria
End With
.Parameters.Add(Pr)
' Create data adapter to use its Fill() method
dda = DbProviderFactories.GetFactory(.Connection).CreateDataAdapter()
' Assign the prepared DbCommand as a select method for the adapter
dda.SelectCommand = Cmd
' A single resultset is expected here
dda.Fill(SearchObjectsBase)
End With
' Set error vars and get rid of it
Call MyBase.SetErrorOutput(Cmd)
' Check for errors and, if any, discard the dataset
If MyBase.ErrorNumber <> 0 Then SearchObjectsBase.Clear()
End Function
I use .NET 4.5, which has a very nice method to automatically select the most appropriate data adapter based on the actual connection.
And here is a call of this function:
Dim XDoc As New XElement("Criteria"), DS As DataSet = Nothing, DT As DataTable
...
DS = .SearchPatients(XDoc.ToString(SaveOptions.None))
' Assign datasource to a grid
Me.dgr_Search.DataSource = DS.Tables.Item(0)
Here, SearchPatients() is a wrapper on top of the SearchObjectsBase().

Rollback the inner transaction of nested transaction

suppose I have following sql statement in sql server 2008:
BEGIN TRANSACTION
SqlStatement1
EXEC sp1
SqlStatement3
COMMIT TRANSACTION
The code of sp1
BEGIN TRANSACTION
SqlStatement2
ROLLBACK TRANSACTION
My question is: Is SqlStatement3 actually executed?
SQL Server doesn't really support nested transactions. There is only one transaction at a time.
This one transaction has a basic nested transaction counter, ##TRANCOUNT. Each consecutive begin transaction increments the counter by one, each commit transaction reduces it by one. Only the commit that reduces the counter to 0 really commits the one transaction.
A rollback transaction undoes the one transaction and clears ##TRANCOUNT.
In your case, the funny result is that SqlStatement3 is run outside a transaction! Your final commit will throw an "The COMMIT TRANSACTION request has no corresponding BEGIN TRANSACTION" exception, but the effects of SqlStatement3 are permanent.
For example:
create table #t (col1 int)
insert #t (col1) values (1)
BEGIN TRANSACTION
update #t set col1 = 2 -- This gets rolled back
BEGIN TRANSACTION
update #t set col1 = 3 -- This gets rolled back too
ROLLBACK TRANSACTION
update #t set col1 = 4 -- This is run OUTSIDE a transaction!
COMMIT TRANSACTION -- Throws error
select col1 from #t
Prints 4. Really. :)
You can use transaction savepoints. sp1 can use a pattern like the one described in Error Handling and Nested Transactions:
create procedure [usp_my_procedure_name]
as
begin
set nocount on;
declare #trancount int;
set #trancount = ##trancount;
begin try
if #trancount = 0
begin transaction
else
save transaction usp_my_procedure_name;
-- Do the actual work here
lbexit:
if #trancount = 0
commit;
end try
begin catch
declare #error int, #message varchar(4000), #xstate int;
select #error = ERROR_NUMBER(), #message = ERROR_MESSAGE(), #xstate = XACT_STATE();
if #xstate = -1
rollback;
if #xstate = 1 and #trancount = 0
rollback
if #xstate = 1 and #trancount > 0
rollback transaction usp_my_procedure_name;
raiserror ('usp_my_procedure_name: %d: %s', 16, 1, #error, #message) ;
end catch
end
Such a pattern allow for the work done in sp1 to rollback, but keep the encompassing transaction active.
Nested transactions can be used. To only rollback the inner transaction, use a savepoint and rollback to the savepoint. In case the inner transaction does not whether it is nested or not, IF statements can be used to find out whether to set a savepoint and whether to rollback or to rollback to a savepoint:
BEGIN TRAN
DECLARE #WILL_BE_NESTED_TRANSACTION BIT = CASE WHEN (##TRANCOUNT > 0) THEN 1 ELSE 0 END
IF #WILL_BE_NESTED_TRANSACTION = 1
SAVE TRAN tran_save
BEGIN TRAN
-- do stuff
IF #WILL_BE_NESTED_TRANSACTION = 1
ROLLBACK TRAN tran_save
ELSE
ROLLBACK
ROLLBACK
Rollback transaction on its own rolls back all transactions.
http://msdn.microsoft.com/en-us/library/ms181299(v=sql.100).aspx
The statement will still be executed - try this
create table #t (i int)
insert #t values (1) -- t contains (1)
begin tran
update #t set i = i +1
select * from #t -- t contains (2)
begin tran
update #t set i = i +1
select * from #t -- t contains (3)
rollback tran -- transaction is rolled back
select * from #t -- t contains (1)
update #t set i = i +1
select * from #t -- t contains (2)
commit -- error occurs
select * from #t -- t contains (2)
drop table #t
Committing inner transactions is ignored by the SQL Server Database Engine. The transaction is either committed or rolled back based on the action taken at the end of the outermost transaction. If the outer transaction is committed, the inner nested transactions are also committed. If the outer transaction is rolled back, then all inner transactions are also rolled back, regardless of whether or not the inner transactions were individually committed.
Nesting Transactions in Microsoft TechNet

Is this good writen transaction in stored procedure

This is the first time that I use transactions and I just wonder am I make this right. Should I change something?
I insert post(wisp). When insert post I need to generate ID in commentableEntity table and insert that ID in wisp table.
ALTER PROCEDURE [dbo].[sp_CreateWisp]
#m_UserId uniqueidentifier,
#m_WispTypeId int,
#m_CreatedOnDate datetime,
#m_PrivacyTypeId int,
#m_WispText nvarchar(200)
AS
BEGIN TRANSACTION
DECLARE #wispId int
INSERT INTO dbo.tbl_Wisps
(UserId,WispTypeId,CreatedOnDate,PrivacyTypeId,WispText)
VALUES
(#m_UserId,#m_WispTypeId,#m_CreatedOnDate,#m_PrivacyTypeId,#m_WispText)
if ##ERROR <> 0
BEGIN
ROLLBACK
RAISERROR ('Error in adding new wisp.', 16, 1)
RETURN
END
SELECT #wispId = SCOPE_IDENTITY()
INSERT INTO dbo.tbl_CommentableEntity
(ItemId)
VALUES
(#wispId)
if ##ERROR <> 0
BEGIN
ROLLBACK
RAISERROR ('Error in adding commentable entity.', 16, 1)
RETURN
END
DECLARE #ceid int
select #ceid = SCOPE_IDENTITY()
UPDATE dbo.tbl_Wisps SET CommentableEntityId = #ceid WHERE WispId = #wispId
if ##ERROR <> 0
BEGIN
ROLLBACK
RAISERROR ('Error in adding wisp commentable entity id.', 16, 1)
RETURN
END
COMMIT
Using try/catch based on #gbn answer:
ALTER PROCEDURE [dbo].[sp_CreateWisp]
#m_UserId uniqueidentifier,
#m_WispTypeId int,
#m_CreatedOnDate datetime,
#m_PrivacyTypeId int,
#m_WispText nvarchar(200)
AS
SET XACT_ABORT, NOCOUNT ON
DECLARE #starttrancount int
BEGIN TRY
SELECT #starttrancount = ##TRANCOUNT
IF #starttrancount = 0
BEGIN TRANSACTION
DECLARE #wispId int
INSERT INTO dbo.tbl_Wisps
(UserId,WispTypeId,CreatedOnDate,PrivacyTypeId,WispText)
VALUES
(#m_UserId,#m_WispTypeId,#m_CreatedOnDate,#m_PrivacyTypeId,#m_WispText)
SELECT #wispId = SCOPE_IDENTITY()
INSERT INTO dbo.tbl_CommentableEntity
(ItemId)
VALUES
(#wispId)
DECLARE #ceid int
select #ceid = SCOPE_IDENTITY()
UPDATE dbo.tbl_Wisps SET CommentableEntityId = #ceid WHERE WispId = #wispId
IF #starttrancount = 0
COMMIT TRANSACTION
END TRY
BEGIN CATCH
IF XACT_STATE() <> 0 AND #starttrancount = 0
ROLLBACK TRANSACTION
RAISERROR ('Error in adding new wisp', 16, 1)
END CATCH
GO
You'd use TRY/CATCH since SQL Server 2005+
Your rollback goes into the CATCH block but your code looks good otherwise (using SCOPE_IDENTITY() etc). I'd also use SET XACT_ABORT, NOCOUNT ON
This is my template: Nested stored procedures containing TRY CATCH ROLLBACK pattern?
Edit:
This allows for nested transactions as per DeveloperX's answer
This template also allows for higher level transactions as per Randy's comment
i think its not good all the time ,but if you want to use more than one stored procedure same time its not good be cause each stored procedure handles the transaction independently
but in this case,you should use try catch block , for exception handling , and preventing keeping transaction open on when an exception raising
I've never considered it a good idea to put transactions in a stored procedure. I think it's much better to start a transaction at a higher level so that can better coordinate multiple database (e.g. stored procedure) calls and treat them all as a single transaction.

Resources