Deadlock error MERGE

Question

Deadlock error MERGE

javirsantos

Mr or Mrs. 500

Points: 579
More actions
September 28, 2016 at 6:37 am

#313947

I have a step that produces the xml and the MERGE table 1 insert and update, if you run it once
from the application to operate normally but if you run several times simultaneously
application from a session ends well and the rest with deadlocks.
The procedure is inside a TRANSACTION from source code, used EF5.0 and Sql 2014 SP1 CU1.
The procedure is as follows:
ALTER PROCEDURE [dbo].[procedure]
@var1 int,
@var2 xml,
@var3 int
AS
BEGIN
SET NOCOUNT ON
MERGE TABLA AS destino
USING(
SELECT
CampoTabla1 = @var1,
CampoTabla2 = T.Item.value ('@V', 'char(2)')
FROM @var2.nodes('/L/I') AS T(Item)
)AS origen ON
origen.Campo1 = destino.Campo1 AND
origen.Campo2 = destino.Campo2
WHEN MATCHED THEN
UPDATE SET
destino.Campo3 = @var3,
destino.Campo4 = 0
WHEN NOT MATCHED BY SOURCE AND
destino.Campo1 = @var1
AND destino.Campo3 = 0 THEN
UPDATE SET
destino.Campo3 = @var3,
destino.Campo4 = GETDATE()
WHEN NOT MATCHED THEN
INSERT
(Campo1,
Campo2,
Campo3,
Campo4,
Campo5)
VALUES
(origen.Campo1,
origen.Campo2,
@var3,
GETDATE(),
0);
END
I modified the MAXDOP parameter of the instance but still the same error.
THE BBDD was read commited snapshot.
The level of parallelism is at 0 (Unlimited)
The MAXDOP is at 8, since the machine has more than 8 cores.
How can I fix it, something happens to them.
Greetings.

Viewing 15 posts - 1 through 15 (of 15 total)

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

Phil Parkin SSC Guru Points: 246985 More actions · Answer 1

I'd suggest that you try adding a merge hint to your MERGE. Start with HOLDLOCK and test ... moving up to TABLOCK if necessary.

You should also take steps to ensure that the MERGE is performing as optimally as possible.

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

spaghettidba SSC Guru Points: 105732 More actions · Answer 2

MERGE is a widely known deadlock generator: https://www.mssqltips.com/sqlservertip/3074/use-caution-with-sql-servers-merge-statement/

-- Gianluca Sartori

javirsantos Mr or Mrs. 500 Points: 579 More actions · Answer 3

According to the page that is attached is a known bug, I have seen that there is a hotfix that you can apply

but the version table only shows up to 201 SQL, I have a sql 2014 with SP1 CU1.

This procedure was before in a SQL 2008R2 SP2 CU1 engine and not failed or not we had

produced the error, the application is the same.

Can you think of any solution to avoid having to modify the procedure.

I only put one example, but with this extructura have 2000 Procedures.

spaghettidba SSC Guru Points: 105732 More actions · Answer 4

You're not running the latest build for SQL 2014. You're missing SP2 and CU1. Maybe the fix you're referring to is included in one of those.

I would test in a test environment to see if applying SP2 and CU1 help.

-- Gianluca Sartori

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 5

You may want to incorporate deadlock handling into your code, perhaps using something like the following snippet I wrote to handle deadlock processing in one of our procedures:

-- The following snippet is for building deadlock handling.

declare @ErrorNumber int,

@ErrorLine int,

@ErrorMessage nvarchar(4000),

@ErrorProcedure nvarchar(128),

@ErrorSeverity int,

@ErrorState int;

while 1 = 1

begin

begin transaction;

begin try;

-- <-- code goes here -->

commit;

break;

end try

begin catch

rollback;

select

@ErrorNumber = ERROR_NUMBER(),

@ErrorLine = ERROR_LINE(),

@ErrorMessage = ERROR_MESSAGE(),

@ErrorProcedure = ERROR_PROCEDURE(),

@ErrorSeverity = ERROR_SEVERITY(),

@ErrorState = ERROR_STATE();

if @ErrorNumber = 1205 -- Trap deadlock

begin

WAITFOR DELAY '00:00:00.05' -- Wait for 50 ms

continue;

end

else begin

-- All other errors

raiserror(@ErrorMessage,@ErrorSeverity,@ErrorState);

end

end catch

end

javirsantos Mr or Mrs. 500 Points: 579 More actions · Answer 6

I tested with tests and test HOLDLOCK With ... moving up to TABLOCK, but the error is still playing.

This occurs when multiple concurrent users try to update.

javirsantos Mr or Mrs. 500 Points: 579 More actions · Answer 7

I set up a test environment with S2 CU1 patch level and gives the same error, deadlock.

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 8

javirsantos (9/30/2016)
I set up a test environment with S2 CU1 patch level and gives the same error, deadlock.

There's a strong relationship between transaction duration and deadlock occurrence. Speeding up your transaction will reduce the likelihood of a deadlock.

Check that the permanent table has an index which will support the update, and consider resolving the xml into a #temp table or @table variable as a separate step.

Posting an actual (as opposed to estimated) execution plan as a .sqlplan attachment will assist folks in tuning the query.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

javirsantos Mr or Mrs. 500 Points: 579 More actions · Answer 9

The XML file I have gotten into a @xml and #XML and the deadlock is still playing.

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 10

javirsantos (10/5/2016)
The XML file I have gotten into a @xml and #XML and the deadlock is still playing.

The next step would be to post an execution plan here, as a .sqlplan attachment.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden