Trace Flag 7470

Trace flag 7470 works in SQL 2017! I'm always a little reluctant to use trace flags, especially after reading that Trace flags were becoming redundant due to more database level settings.

Anyway, I've eliminated the Sort operator spilling to tempdb by applying 7470.

Note, the documentation says it doesn't apply to SQL 2017, but I had it on good authority that it would,

Trace Flag 7470 : Fix operator spills to tempdb

TSQL : HEAP table usage script

This script reports usage of HEAP tables over a 15 minute period.
It reports reads, writes and row counts for the tables.
I wrote it as I found a system exhibiting high values for Performance Counter 'forwarded records/sec'.

See Also : TSQL : HEAPS and forwarded records

IF OBJECT_ID('tempdb..#Snap1') IS NOT NULL DROP TABLE #Snap1
IF OBJECT_ID('tempdb..#Snap2') IS NOT NULL DROP TABLE #Snap2
IF OBJECT_ID('tempdb..#rowcounts') IS NOT NULL DROP TABLE #rowcounts

SELECT   
      s.object_id
    , OBJECT_SCHEMA_NAME(s.object_id) AS schemaname
    , OBJECT_NAME(s.object_id) AS tablename
    , i.name AS indexname
       , SUM(user_seeks) user_seeks, SUM(user_scans) user_scans, SUM(user_lookups) user_lookups, reads = SUM(user_seeks + user_scans + user_lookups), writes =  SUM(user_updates)
INTO #snap1
FROM sys.dm_db_index_usage_stats AS s
INNER JOIN sys.indexes AS i
ON s.object_id = i.object_id
AND i.index_id = s.index_id
WHERE OBJECTPROPERTY(s.object_id,'IsUserTable') = 1
  AND OBJECTPROPERTY(s.object_id,'TableHasClustIndex') = 0 -- Reporting only on HEAPS
GROUP BY s.object_id,i.name

WAITFOR DELAY '00:15:00'

SELECT   
      s.object_id
    , OBJECT_SCHEMA_NAME(s.object_id) AS schemaname
    , OBJECT_NAME(s.object_id) AS tablename
    , i.name AS indexname
       , SUM(user_seeks) user_seeks, SUM(user_scans) user_scans, SUM(user_lookups) user_lookups, reads = SUM(user_seeks + user_scans + user_lookups), writes =  SUM(user_updates)
INTO #snap2
FROM sys.dm_db_index_usage_stats AS s
INNER JOIN sys.indexes AS i
ON s.object_id = i.object_id
AND i.index_id = s.index_id
WHERE OBJECTPROPERTY(s.object_id,'IsUserTable') = 1
  AND OBJECTPROPERTY(s.object_id,'TableHasClustIndex') = 0 -- Reporting only on HEAPS
GROUP BY s.object_id,i.name

SELECT   sys_schemas.name AS schemaName
  ,sys_objects.name AS tableName
  ,sys_indexes.name AS indexName
  ,SUM(partition_stats.used_page_count) * 8 AS indexsizekb
  ,MAX(partitions.rows) AS tablerows
INTO #rowcounts
FROM sys.dm_db_partition_stats partition_stats
INNER JOIN sys.indexes sys_indexes
  ON partition_stats.[object_id] = sys_indexes.[object_id] 
    AND partition_stats.index_id = sys_indexes.index_id
INNER JOIN sys.objects sys_objects
  ON sys_objects.[object_id] = partition_stats.[object_id] 
INNER JOIN sys.schemas sys_schemas  
  ON sys_objects.[schema_id] = sys_schemas.[schema_id] 
  AND sys_schemas.name <> 'SYS'
INNER JOIN sys.partitions partitions
  ON partitions.object_id = partition_stats.object_id 
WHERE OBJECTPROPERTY(sys_objects.object_id,'TableHasClustIndex') = 0
GROUP BY sys_schemas.name
  ,sys_objects.name
  ,sys_indexes.name
ORDER BY sys_schemas.name
  ,sys_objects.name
  ,sys_indexes.name

SELECT   i.schemaname
  ,i.tablename
  ,i.indexname
  ,(j.user_seeks - i.user_seeks)  AS seeks_period
  ,(i.user_scans - i.user_scans) AS scans_period
  ,(j.user_lookups - i.user_lookups) AS lookups_period
  ,(j.reads - i.reads) AS reads_period
  ,(j.writes - i.writes) AS writes_period
  ,r.tablerows
  ,r.indexsizekb
FROM #Snap1 i
INNER JOIN #Snap2 j ON i.tablename = j.tablename
INNER JOIN #rowcounts r ON r.tableName = i.tablename AND r.indexName = i.indexname
AND COALESCE(i.indexname,'!') = COALESCE(j.indexname,'!')

TSQL : HEAPS and Forwarded Records

A little bemused that I'm still finding this in application code in the 21st Century, but there you go. A live app experiencing performance problems due to a significant number of heaps, each frequently used.

SELECT OBJECT_NAME(stats.object_id) as [Table],
idx.name as [Index],
stats.index_type_desc,
stats.page_count,
stats.avg_fragmentation_in_percent,
stats.forwarded_record_count
FROM sys.dm_db_index_physical_stats (DB_ID(), NULL, NULL, NULL, 'DETAILED') AS stats
INNER JOIN sys.indexes AS idx (NOLOCK)
ON stats.OBJECT_ID = idx.OBJECT_ID
AND stats.index_id = idx.index_id
WHERE forwarded_record_count > 0
ORDER BY forwarded_record_count desc

See also : HEAP table usage script

TSQL : Clustered Indexes, Identities & Primary Keys

I'm Currently looking at a large database where there are

• HEAPS (tables without a clustered index)
• IDENTITY columns that could serve as Primary Keys (but aren't)
• PRIMARY KEYS that have been defined as non-clustered (someone has scripted them out and ignored the default clustered status of a PK).

Anyway, this is coming in useful ...

SELECT   SCHEMA_NAME(schema_id) AS SchemaName
  ,name AS TableName
  ,OBJECTPROPERTY(object_id,'TableHasClustIndex') HasClusteredIndex
  ,OBJECTPROPERTY(object_id,'TableHasIdentity') HasIdentity
  ,OBJECTPROPERTY(object_id,'TableHasprimarykey') HasPrimaryKey
FROM sys.tables
WHERE OBJECTPROPERTY(object_id,'TableHasClustIndex') = 0
ORDER BY 1, 2

SQLTips : CONCAT

The CONCAT function is available from SQL 2016+. It can save you a lot of messing around with adding strings together and coping with NULLS and empty strings.

SELECT CONCAT('The cat sat',' ', 'on', ' ', 'the mat')
GO
 

SELECT CONCAT(NULL, ' and void')
GO

 
DECLARE @int int = 99
DECLARE @varchar varchar(13) = ' red balloons'
SELECT CONCAT(@int,@varchar)
GO

 
DECLARE @varchar varchar(20) = 'The date today is '
DECLARE @dt datetime = GETDATE()
SELECT CONCAT(@varchar,@dt)
GO

sp_ms_marksystemobject

On searching for a table I found it to be hidden. Turns out it was marked as 'ms shipped' i.e was a system object.

select * from sys.tables WHERE name = 'sysssislog'

To achieve this yourself you can use sp_ms_marksystemobject

EXEC sp_ms_marksystemobject 'dbo.sysssislog'

Tables with Computed Columns

Tables with Computed Columns

SELECT
  SCHEMA_NAME(schema_id) AS schemaname
, t.name as tablename
, cc.name
, cc.definition
FROM sys.tables t
INNER JOIN sys.Computed_columns cc
ON t.object_id = cc.object_id

and without

select
  SCHEMA_NAME(schema_id) AS schemaname
, t.name as tablename
from  sys.tables t
WHERE object_id not in (SELECT object_id FROM sys.Computed_columns)