Temporary tables and constant statement recompilation - sql-server

I am stress testing a system that is using a temporary table in dynamic SQL. The table is created early on in the transaction and is filled by several dynamic SQL statements in several stored procedures that are executed as part of the batch using statements of the form:
INSERT #MyTable (...)
SELECT ...
where the SELECT statement is reasonably complicated in that it may contain UNION ALL and UNPIVOT statements and refer to several UDFs. All strings are executed using sp_executesql and Parameter Sniffing is enabled.
I have noticed that under load I am seeing a lot of RESOURCE_SEMAPHORE_QUERY_COMPILE waits where the query text being recompiled is present and identical in several waits at the same time and appears throughout the stress test which lasts about 5mins. The memory consumption on the server usually sits around 60% utilization and there is no limit on how much SQL Server can consume. The limiting factor appears to be CPU, which is constantly at >95% during the test.
I have profiled the server during the test to observe the SQL:StmtRecompile event which highlights the reason for the recompile is:
5 - Temp table changed
but the temp table is the same every time and there are no DDL statements performed against the table once it has been created, apart from when it is dropped at the end of the batch.
So far, I have tried:
Enabling the "optimize for ad hoc workloads" option
OPTION(KEEPFIXED PLAN)
Changing the dynamic statement to just the SELECT and then using INSERT ... EXEC so the temp table is not in the executed string
All of these have made no difference and the waits persist.
Why would SQL think that is needs to recompile these identical queries each time they are executed and how can I get it to keep and reuse the cached plans it is creating?
Note: I cannot change the temp table to an In-Memory table because sometimes the stored procedures using this may have to query another database on the same instance.
This is using SQL Server 2016 SP1 CU7.

It appears that removing the insertion into the temp table in the dynamic SQL strings improves performance significantly. For example, changing this:
EXEC sp_executesql
N'INSERT #tempTable (...) SELECT ... FROM ...'
where the SELECT statement is non-trivial, to this:
INSERT #tempTable (...)
EXEC sp_executesql
N'SELECT ... FROM ...'
greatly reduced the number of blocks created during compilation. Unfortunately, recompilation of the queries is not avoided it's just that the queries being recompiled are now much simpler and therefore less CPU intensive.
I have also found it more performant to create an In-Memory Table Type with the same columns as the temp table, perform the complex insertions into a table variable of that type and perform a single insert from the table variable into the temp table at the end.

Related

Way to get the CREATE statement of global temporary table?

I have stored procedures I need to rewrite that used global temporary tables but now must use local temporary tables so that there is no cross over between users/the prod and dev db which are on the same SQL Server instances or any other new databases this database will be share the instance with.
I have to convert a lot of code that is like this
EXEC sp_executesql
N'SELECT manyDynamicallyCreatedColumns INTO ##someTempTable
I now for this to work for local tables I must instead create the table first and then INSERT INTO like
CREATE TABLE #tempTable (manyColumns);
EXEC sp_executesql
N'INSERT INTO #tempTable SELECT manyColumns FROM somewhere'
The issue I face is that this stored procedure has 8 different scenarios that lead to the global temp table. The select statement of the SQL is dynamically generated via a number of other queries. I think what would be the easiest way for me to figure out what my CREATE TABLE #tempTable should look like is if I could print/select what global temporary table looks like after it is made, for each of these 8 scenarios. Then I would just need a If/Else If statement that creates the local table appropriately before proceeding. But I don't know how or if this is possible.
For real tables I can right click -> Script to...-> Create. I don't know if there is an analogous way to do this via scripting that works for global temporary tables. Is there?

conditional stored procedure performance vs. code build statement

currenty i am working on a report system for our data archive.
the aim is to select data for every 1st of a month, every full hour and so on.
So I have a bunch of parameters to select the data down to a single hour.
To achieve that I used CASE statements to adjust the select like this:
SELECT
MIN(cd.Timestamp) as Mintime,
--Hours
CASE
WHEN
#SelHour IS NOT NULL
THEN
DATEPART(HOUR, cd.Timestamp)
END as Hour,
... -- more CASES up to DATEPART(YEAR, cd.Timestamp)
FROM dbo.CustomerData cd
... -- filter data and other stuff
This statements works good for me so far, but I am a bit worried about the performance of the stored procedure. Because I don't know how the server will behave with this "changing" statement. The result can vary between a 20 row result up to a 250.000 rows and more. Depending on the given parameters. As far as I know the sql server saves the query plan and reuses it for future execution.
When it saves the plan for the 20 row result the performance for the 250.000 result is propably pretty poor.
Now I am wondering whats the better aproach. Using this stored procedure or create the statement inside my c# backend and pass the "adjusted" statement to the sql server?
Thanks and greetings
For 20 rows result set it will work good anywhere. But for returning 250k records to c# code seems change in design for this code since loading 250k records in memory & looping will also consume significant memory and such concurrent requests from different session/user will multiply load exponentially.
Anyway to address problem with SQL Server reusing same query plan, you can recompile query plans selectively or every time. These are options available for Recompile execution plan:
OPTION(RECOMPILE)
SELECT
MIN(cd.Timestamp) as Mintime,
--Hours
CASE
WHEN
#SelHour IS NOT NULL
THEN
DATEPART(HOUR, cd.Timestamp)
END as Hour,
... -- more CASES up to DATEPART(YEAR, cd.Timestamp)
FROM dbo.CustomerData cd
... -- filter data and other stuff
OPTION(RECOMPILE)
WITH RECOMPILE Option this will recompile execution plan every time
CREATE PROCEDURE dbo.uspStoredPrcName
#ParamName varchar(30) = 'abc'
WITH RECOMPILE
AS
...
RECOMPILE Query Hint providing WITH RECOMPILE in execute
NOTE: this will require CREATE PROCEDURE permission in the database and ALTER permission on the schema in which the procedure is being created.
EXECUTE uspStoredPrcName WITH RECOMPILE;
GO
sp_recompile System Stored Procedure
NOTE: Requires ALTER permission on the specified procedure.
EXEC sp_recompile N'dbo.uspStoredPrcName ';
GO
For more details on Recompile refer Microsoft Docs:
https://learn.microsoft.com/en-us/sql/relational-databases/stored-procedures/recompile-a-stored-procedure?view=sql-server-ver15

Change table variable to temp table

I have a certain stored procedure written years ago. It uses a table variable. Inside the code it's doing:
declare #OpportunityVS3 as SF_OpportunityMerge
insert #TableVariable select * from /*...*/
This is inside a job that runs every five minutes. There used to be no issues in the past because the select used to collect around 10 to 1000 records… But now sometimes the select is trying to insert a million records.
Given this issue, I think I am forced to change the #TableVariable to a #TempTable. What do you think? Do I have any other option?
At the end of the SP, it’s passing the Variable to another SP:
exec [dbo].[SF_MergeOpportunity] #OpportunityVS3, #LastUpdateDate
I guess I will have to write it all in the same stored procedure, because temporary tables cannot be passed, right?
Below is the definition of the #TableVariable that is being used and the code I would have to change:
declare #OpportunityVS3 as SF_OpportunityMerge
insert #OpportunityVS3 select Opportunity_id, Salesforce_id, AccountId, Age__c,
-- (continues)
merge [BU2].[dbo].[Salesforce_Opportunity] as TARGET
using ( select Opportunity_id,
Salesforce_id,
AccountId,
Age__c,
Amount,
Bill99Amount__c,
BJ_Marketing__c
-- (continues)
from #OpportunityVS3 ) as SOURCE on (TARGET.Opportunity_id = SOURCE.Opportunity_id)
exec [dbo].[SF_MergeOpportunity] #OpportunityVS3, #LastUpdateDate
[![enter image description here][1]][1]
[![enter image description here][2]][2]
What you can try without changing the way you work:
Force the compiler to take cardinalities of TABLE variables into account by specifying OPTION(RECOMPILE) in queries using the TABLE variables
Supply a suitable UNIQUE INDEX or PRIMARY KEY for the TABLE variable
A downside of TABLE variables is that there are no statistics associated with them. The SQL compiler will produce better execution plans when it can take statistics into account. Therefore you can improve performance considerably when you switch to temporary tables because these do have statistics.
When you switch to temporary tables, you would no longer pass a TABLE variable to your stored procedure anymore. Your stored procedure would then be written using a temporary table that it knows exists beforehand - ie was created before the stored procedure is executed.
You would write your stored procedure as though the temporary table were an ordinary table. When you write the stored procedure using a temporary table, SQL Server management studio will underline certain parts in red so it appears as though there are errors in your stored procedure. But if the syntax is correct, creating/altering the stored procedure will work just fine.
Read this excellent essay on the differences between TABLE variables and temporary tables, to see how usage may impact performance. Especially topics No column statistics and Indexes.

Concurrency problems with temp tables in consequent batches?

I have sometimes a problem when running a script. I have the probelm when using an application (that I didn't write and therefore cannot debug) that launches the scripts. This app isn't returning the full error from SQL Server, but just the error description, so I don't know exactly where th error comes.
I have the error only using this tool (it is a tool that sends the queries directly to SQL Server, using a DAC component), if I run the query manuallyin management studio I don't have the error. (This error moreover occurs only on a particular database).
My query is something like:
SELECT * INTO #TEMP_TABLE
FROM ANOTHER_TABLE
GO
--some other commands here
GO
INSERT INTO SOME_OTHER_TABLE(FIELD1,FIELD2)
SELECT FIELDA, FIELDB
FROM #TEMP_TABLE
GO
DROP TABLE #TEMP_TABLE
GO
The error I get is #TEMP_TABLE is not a valid object
So somehow i suspect that the DROP statement is executed before the INSERT statement.
But AFAIK when a GO is there the next statement is not executed until the previous has been completed.
Now I suspoect that this is not true with temp tables... Or do you have another ideas?
Your problem is most likely caused by either an end of session prior to the DROP TABLE causing SQL Server to automatically drop the table or the DROP TABLE is being executed in a different session than the other code (that created and used the temporary table) causing the table not to be visible.
I am assuming that stored procedures are not involved here, because it looks like you are just executing batches, since local temporary tables are also dropped when a stored proc is exited.
There is a good description of local temporary table behavior in this article on Temporary Tables in SQL Server:
You get housekeeping with Local Temporary tables; they are
automatically dropped when they go out of scope, unless explicitly
dropped by using DROP TABLE. Their scope is more generous than a table
Variable so you don't have problems referencing them within batches or
in dynamic SQL. Local temporary tables are dropped automatically at
the end of the current session or procedure. Dropping it at the end of
the procedure that created it can cause head-scratching: a local
temporary table that is created within a stored procedure or session
is dropped when it is finished so it cannot be referenced by the
process that called the stored procedure that created the table. It
can, however, be referenced by any nested stored procedures executed
by the stored procedure that created the table. If the nested
procedure references a temporary table and two temporary tables with
the same name exist at that time, which table is the query is resolved
against?
I would start up SQL Profiler and verify if your tool uses one connection to execute all batches, or if it disconnects/reconnects. Also it could be using a connection pool.
Anyway, executing SQL batches from a file is so simple that you might develop your own tool very quickly and be better off.

Errors: "INSERT EXEC statement cannot be nested." and "Cannot use the ROLLBACK statement within an INSERT-EXEC statement." How to solve this?

I have three stored procedures Sp1, Sp2 and Sp3.
The first one (Sp1) will execute the second one (Sp2) and save returned data into #tempTB1 and the second one will execute the third one (Sp3) and save data into #tempTB2.
If I execute the Sp2 it will work and it will return me all my data from the Sp3, but the problem is in the Sp1, when I execute it it will display this error:
INSERT EXEC statement cannot be nested
I tried to change the place of execute Sp2 and it display me another error:
Cannot use the ROLLBACK statement
within an INSERT-EXEC statement.
This is a common issue when attempting to 'bubble' up data from a chain of stored procedures. A restriction in SQL Server is you can only have one INSERT-EXEC active at a time. I recommend looking at How to Share Data Between Stored Procedures which is a very thorough article on patterns to work around this type of problem.
For example a work around could be to turn Sp3 into a Table-valued function.
This is the only "simple" way to do this in SQL Server without some giant convoluted created function or executed sql string call, both of which are terrible solutions:
create a temp table
openrowset your stored procedure data into it
EXAMPLE:
INSERT INTO #YOUR_TEMP_TABLE
SELECT * FROM OPENROWSET ('SQLOLEDB','Server=(local);TRUSTED_CONNECTION=YES;','set fmtonly off EXEC [ServerName].dbo.[StoredProcedureName] 1,2,3')
Note: You MUST use 'set fmtonly off', AND you CANNOT add dynamic sql to this either inside the openrowset call, either for the string containing your stored procedure parameters or for the table name. Thats why you have to use a temp table rather than table variables, which would have been better, as it out performs temp table in most cases.
OK, encouraged by jimhark here is an example of the old single hash table approach: -
CREATE PROCEDURE SP3 as
BEGIN
SELECT 1, 'Data1'
UNION ALL
SELECT 2, 'Data2'
END
go
CREATE PROCEDURE SP2 as
BEGIN
if exists (select * from tempdb.dbo.sysobjects o where o.xtype in ('U') and o.id = object_id(N'tempdb..#tmp1'))
INSERT INTO #tmp1
EXEC SP3
else
EXEC SP3
END
go
CREATE PROCEDURE SP1 as
BEGIN
EXEC SP2
END
GO
/*
--I want some data back from SP3
-- Just run the SP1
EXEC SP1
*/
/*
--I want some data back from SP3 into a table to do something useful
--Try run this - get an error - can't nest Execs
if exists (select * from tempdb.dbo.sysobjects o where o.xtype in ('U') and o.id = object_id(N'tempdb..#tmp1'))
DROP TABLE #tmp1
CREATE TABLE #tmp1 (ID INT, Data VARCHAR(20))
INSERT INTO #tmp1
EXEC SP1
*/
/*
--I want some data back from SP3 into a table to do something useful
--However, if we run this single hash temp table it is in scope anyway so
--no need for the exec insert
if exists (select * from tempdb.dbo.sysobjects o where o.xtype in ('U') and o.id = object_id(N'tempdb..#tmp1'))
DROP TABLE #tmp1
CREATE TABLE #tmp1 (ID INT, Data VARCHAR(20))
EXEC SP1
SELECT * FROM #tmp1
*/
My work around for this problem has always been to use the principle that single hash temp tables are in scope to any called procs. So, I have an option switch in the proc parameters (default set to off). If this is switched on, the called proc will insert the results into the temp table created in the calling proc. I think in the past I have taken it a step further and put some code in the called proc to check if the single hash table exists in scope, if it does then insert the code, otherwise return the result set. Seems to work well - best way of passing large data sets between procs.
This trick works for me.
You don't have this problem on remote server, because on remote server, the last insert command waits for the result of previous command to execute. It's not the case on same server.
Profit that situation for a workaround.
If you have the right permission to create a Linked Server, do it.
Create the same server as linked server.
in SSMS, log into your server
go to "Server Object
Right Click on "Linked Servers", then "New Linked Server"
on the dialog, give any name of your linked server : eg: THISSERVER
server type is "Other data source"
Provider : Microsoft OLE DB Provider for SQL server
Data source: your IP, it can be also just a dot (.), because it's localhost
Go to the tab "Security" and choose the 3rd one "Be made using the login's current security context"
You can edit the server options (3rd tab) if you want
Press OK, your linked server is created
now your Sql command in the SP1 is
insert into #myTempTable
exec THISSERVER.MY_DATABASE_NAME.MY_SCHEMA.SP2
Believe me, it works even you have dynamic insert in SP2
I found a work around is to convert one of the prods into a table valued function. I realize that is not always possible, and introduces its own limitations. However, I have been able to always find at least one of the procedures a good candidate for this. I like this solution, because it doesn't introduce any "hacks" to the solution.
I encountered this issue when trying to import the results of a Stored Proc into a temp table, and that Stored Proc inserted into a temp table as part of its own operation. The issue being that SQL Server does not allow the same process to write to two different temp tables at the same time.
The accepted OPENROWSET answer works fine, but I needed to avoid using any Dynamic SQL or an external OLE provider in my process, so I went a different route.
One easy workaround I found was to change the temporary table in my stored procedure to a table variable. It works exactly the same as it did with a temp table, but no longer conflicts with my other temp table insert.
Just to head off the comment I know that a few of you are about to write, warning me off Table Variables as performance killers... All I can say to you is that in 2020 it pays dividends not to be afraid of Table Variables. If this was 2008 and my Database was hosted on a server with 16GB RAM and running off 5400RPM HDDs, I might agree with you. But it's 2020 and I have an SSD array as my primary storage and hundreds of gigs of RAM. I could load my entire company's database to a table variable and still have plenty of RAM to spare.
Table Variables are back on the menu!
I recommend to read this entire article. Below is the most relevant section of that article that addresses your question:
Rollback and Error Handling is Difficult
In my articles on Error and Transaction Handling in SQL Server, I suggest that you should always have an error handler like
BEGIN CATCH
IF ##trancount > 0 ROLLBACK TRANSACTION
EXEC error_handler_sp
RETURN 55555
END CATCH
The idea is that even if you do not start a transaction in the procedure, you should always include a ROLLBACK, because if you were not able to fulfil your contract, the transaction is not valid.
Unfortunately, this does not work well with INSERT-EXEC. If the called procedure executes a ROLLBACK statement, this happens:
Msg 3915, Level 16, State 0, Procedure SalesByStore, Line 9 Cannot use the ROLLBACK statement within an INSERT-EXEC statement.
The execution of the stored procedure is aborted. If there is no CATCH handler anywhere, the entire batch is aborted, and the transaction is rolled back. If the INSERT-EXEC is inside TRY-CATCH, that CATCH handler will fire, but the transaction is doomed, that is, you must roll it back. The net effect is that the rollback is achieved as requested, but the original error message that triggered the rollback is lost. That may seem like a small thing, but it makes troubleshooting much more difficult, because when you see this error, all you know is that something went wrong, but you don't know what.
I had the same issue and concern over duplicate code in two or more sprocs. I ended up adding an additional attribute for "mode". This allowed common code to exist inside one sproc and the mode directed flow and result set of the sproc.
what about just store the output to the static table ? Like
-- SubProcedure: subProcedureName
---------------------------------
-- Save the value
DELETE lastValue_subProcedureName
INSERT INTO lastValue_subProcedureName (Value)
SELECT #Value
-- Return the value
SELECT #Value
-- Procedure
--------------------------------------------
-- get last value of subProcedureName
SELECT Value FROM lastValue_subProcedureName
its not ideal, but its so simple and you don't need to rewrite everything.
UPDATE:
the previous solution does not work well with parallel queries (async and multiuser accessing) therefore now Iam using temp tables
-- A local temporary table created in a stored procedure is dropped automatically when the stored procedure is finished.
-- The table can be referenced by any nested stored procedures executed by the stored procedure that created the table.
-- The table cannot be referenced by the process that called the stored procedure that created the table.
IF OBJECT_ID('tempdb..#lastValue_spGetData') IS NULL
CREATE TABLE #lastValue_spGetData (Value INT)
-- trigger stored procedure with special silent parameter
EXEC dbo.spGetData 1 --silent mode parameter
nested spGetData stored procedure content
-- Save the output if temporary table exists.
IF OBJECT_ID('tempdb..#lastValue_spGetData') IS NOT NULL
BEGIN
DELETE #lastValue_spGetData
INSERT INTO #lastValue_spGetData(Value)
SELECT Col1 FROM dbo.Table1
END
-- stored procedure return
IF #silentMode = 0
SELECT Col1 FROM dbo.Table1
Declare an output cursor variable to the inner sp :
#c CURSOR VARYING OUTPUT
Then declare a cursor c to the select you want to return.
Then open the cursor.
Then set the reference:
DECLARE c CURSOR LOCAL FAST_FORWARD READ_ONLY FOR
SELECT ...
OPEN c
SET #c = c
DO NOT close or reallocate.
Now call the inner sp from the outer one supplying a cursor parameter like:
exec sp_abc a,b,c,, #cOUT OUTPUT
Once the inner sp executes, your #cOUT is ready to fetch. Loop and then close and deallocate.
If you are able to use other associated technologies such as C#, I suggest using the built in SQL command with Transaction parameter.
var sqlCommand = new SqlCommand(commandText, null, transaction);
I've created a simple Console App that demonstrates this ability which can be found here:
https://github.com/hecked12/SQL-Transaction-Using-C-Sharp
In short, C# allows you to overcome this limitation where you can inspect the output of each stored procedure and use that output however you like, for example you can feed it to another stored procedure. If the output is ok, you can commit the transaction, otherwise, you can revert the changes using rollback.
On SQL Server 2008 R2, I had a mismatch in table columns that caused the Rollback error. It went away when I fixed my sqlcmd table variable populated by the insert-exec statement to match that returned by the stored proc. It was missing org_code. In a windows cmd file, it loads result of stored procedure and selects it.
set SQLTXT= declare #resets as table (org_id nvarchar(9), org_code char(4), ^
tin(char9), old_strt_dt char(10), strt_dt char(10)); ^
insert #resets exec rsp_reset; ^
select * from #resets;
sqlcmd -U user -P pass -d database -S server -Q "%SQLTXT%" -o "OrgReport.txt"

Resources