I was recently tasked with debugging a strange problem within an e-commerce application. After an application upgrade the site started to hang from time to time and I was sent in to debug. After checking the event log I found that the SQL-server wrote ~200 000 events in a couple of minutes with the message saying that a constraint had failed. After much debugging and some tracing I found the culprit. I've removed some unnecessary code and cleaned it up a bit but essentially this is it
WHILE EXISTS (SELECT * FROM ShoppingCartItem WHERE ShoppingCartItem.PurchID = @PurchID)
BEGIN
SELECT TOP 1
@TmpGFSID = ShoppingCartItem.GFSID,
@TmpQuantity = ShoppingCartItem.Quantity,
@TmpShoppingCartItemID = ShoppingCartItem.ShoppingCartItemID,
FROM
ShoppingCartItem INNER JOIN GoodsForSale on ShoppingCartItem.GFSID = GoodsForSale.GFSID
WHERE ShoppingCartItem.PurchID = @PurchID
EXEC @ErrorCode = spGoodsForSale_ReverseReservations @TmpGFSID, @TmpQuantity
IF @ErrorCode <> 0
BEGIN
Goto Cleanup
END
DELETE FROM ShoppingCartItem WHERE ShoppingCartItem.ShoppingCartItemID = @TmpShoppingCartItemID
-- @@ROWCOUNT is 1 after this
END
Facts:
The procedure has been rewritten to select the rows that should be deleted into a temporary in-memory table instead so the immediate problem is solved but this really sparked my curiosity.
Why does it loop forever?
Clarification: The delete doesn't fail (@@rowcount is 1 after the delete stmt when debugged) Clarification 2: It shouldn't matter whether or not the SELECT TOP ... clause is ordered by any specific field since the record with the returned id will be deleted so in the next loop it should get another record.
Update: After checking the subversion logs I found the culprit commit that made this stored procedure to go haywire. The only real difference that I can find is that there previously was no join in the SELECT TOP 1 statement i.e. without that join it worked without any transaction statements surrounding the delete. It appears to be the introduction of the join that made SQL server more picky.
Update clarification: brien pointed out that there's no need for the join but we actually do use some fields from the GoodsForSale table but I've removed them to keep the code simply so that we can concentrate on the problem at hand
FROM
ShoppingCartItem
INNER JOIN
GoodsForSale
on ShoppingCartItem.GFSID = GoodsForSale.GFSID
Oops, your join brings the result set down to zero rows.
SELECT TOP 1
@TmpGFSID = ShoppingCartItem.GFSID,
@TmpQuantity = ShoppingCartItem.Quantity,
@TmpShoppingCartItemID =
ShoppingCartItem.ShoppingCartItemID
Oops, you used multi-assignment against a set with no rows. This causes the variables to remain unchanged (they will have the same value that they had last time through the loop). The variables do NOT get assigned to null in this case.
If you put this code at the start of the loop, it will (correctly) fail faster:
SELECT
@TmpGFSID = null,
@TmpQuantity = null,
@TmpShoppingCartItemID = null
If you change your code to fetch a key (without joining) and then fetching the related data by key in a second query, you'll win.
Are you operating in explicit or implicit transaction mode?
Since you're in explicit mode, I think you need to surround the DELETE operation with BEGIN TRANSACTION and COMMIT TRANSACTION statements.
WHILE EXISTS (SELECT * FROM ShoppingCartItem WHERE ShoppingCartItem.PurchID = @PurchID)
BEGIN
SELECT TOP 1
@TmpGFSID = ShoppingCartItem.GFSID,
@TmpQuantity = ShoppingCartItem.Quantity,
@TmpShoppingCartItemID = ShoppingCartItem.ShoppingCartItemID,
FROM
ShoppingCartItem INNER JOIN GoodsForSale on ShoppingCartItem.GFSID = GoodsForSale.GFSID
WHERE ShoppingCartItem.PurchID = @PurchID
EXEC @ErrorCode = spGoodsForSale_ReverseReservations @TmpGFSID, @TmpQuantity
IF @ErrorCode <> 0
BEGIN
Goto Cleanup
END
BEGIN TRANSACTION delete
DELETE FROM ShoppingCartItem WHERE ShoppingCartItem.ShoppingCartItemID = @TmpShoppingCartItemID
-- @@ROWCOUNT is 1 after this
COMMIT TRANSACTION delete
END
Clarification: The reason you'd need to use transactions is that the delete doesn't actually happen in the database until you do a COMMIT operation. This is generally used when you have multiple write operations in an atomic transaction. Basically, you only want the changes to happen to the DB if all of the operations are successful.
In your case, there's only 1 operation, but since you're in explicit transaction mode, you need to tell SQL Server to really make the changes.
Is there a record in ShoppingCartItem with that @PurchID where the GFSID is not in the GoodsForSale table? That would explain why the EXISTS returns true, but there are no more records to delete.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With