web-dev-qa-db-ja.com

SQL Serverのインデックスと統計のメンテナンスのパフォーマンスに関する質問

非常に多数のインデックス（> 250.000）を持つデータベースでIndexOptimizeストアドプロシージャを使用しようとしました。 @Indexesパラメータを設定して作業を絞り込む場合でも、ストアドプロシージャが処理する必要のあるデータを収集する最初のステップには数時間かかります。

SQL Server Maintenance Solutionバージョン：2019-02-10 10:40:47最新のCU14がインストールされたSQL Server 2017 Standard Edition。

ある顧客で、500.000を超えるインデックスを持つデータベースを見ました。 12時間後、データ収集ステップはまだ実行されていました。

@Indexesを単一のインデックスに設定すると、すぐに実行が開始されるはずです。

これが私のストアドプロシージャコールの例です。

EXECUTE dbo.IndexOptimize
@Databases = 'db',
@Indexes =  'db.dbo.'
@FragmentationLow = NULL,
@FragmentationMedium = 'INDEX_REORGANIZE,INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationHigh = 'INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationLevel1 = 30,
@FragmentationLevel2 = 50,
@UpdateStatistics = 'ALL',
@OnlyModifiedStatistics = 'Y',
@LogToTable = 'Y',
@LockTimeout = 60

非常に多くのインデックスを持つデータベースにIndexOptimizeを使用した経験を誰かと共有してもらえますか？

sql-serverindexsql-server-2017ola-hallengrenindex-maintenance
4
Joerg Rueschenschmidt

非常に多くのインデックスを持つデータベースにIndexOptimizeを使用した経験を誰かと共有してもらえますか？

そうです、Olaのスクリプトは、最初にすべてのインデックス関連のDMVからデータを取得し、1つのテーブルまたは1つのインデックスが指定されている場合でも、それらを@tmpIndexesStatistics内に挿入します。多くのインデックスの場合、かなり最初のコマンドである最初のコマンドでスタックします。クエリについては以下を参照してください。この問題の考えられる修正については、以下を参照してください。

TL; DR

クエリの1つで、動的管理ビューは、複数のフィルターを使用して数回呼び出されます。

このすべてがかなり大きな実行計画になります。

一時テーブルを作成し、クエリを実行する前にDMVからのデータを保存することは、回避策です。これらの一時テーブル/書き換えのインデックスを使用してさらに最適化できます。

テストデータ：> 100K空のテーブル+> 200k空のインデックス

USE Test2
GO
SET NOCOUNT ON;
DECLARE @SQL NVARCHAR(MAX)
DECLARE @i int = 0

DECLARE @counter int = 1000
WHILE @i <+ @counter
BEGIN
SET @SQL = N'CREATE TABLE dbo.'+QUOTENAME(cast(@i as nvarchar(20)))+'( id int, val varchar(255));'
EXEC(@sql)
SET @SQL = N'CREATE INDEX IX_'+cast(@i as nvarchar(20))+ ' ON '+QUOTENAME(cast(@i as nvarchar(20)))+'(id) '
EXEC(@sql)
SET @SQL = N'CREATE INDEX IX_'+cast(@i as nvarchar(20))+ '_2 ON '+QUOTENAME(cast(@i as nvarchar(20)))+'(val) '
EXEC(@sql)
set @I +=1

IF @i = @counter
BEGIN

  IF @counter < 100000
  BEGIN
        SET @counter += 1000;
  END
END
END

select count(*) from sys.tables;
--100731

select count(*) from sys.indexes  where index_id != 0;
--201614

すべてのインデックスで機能するようにパラメータを変更し、ステートメントを実行する代わりに印刷する場合（@Execute='N'）

EXECUTE MNGDB.dbo.IndexOptimize
@Databases = 'test2',
@Indexes =  'test2.dbo.',
@FragmentationLow = NULL,
@FragmentationMedium = 'INDEX_REORGANIZE,INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationHigh = 'INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationLevel1 = 30,
@FragmentationLevel2 = 50,
@UpdateStatistics = 'ALL',
@OnlyModifiedStatistics = 'Y',
@LogToTable = 'Y',
@LockTimeout = 60,
@Execute='N';

この巨大なクエリが登場します

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED;
USE [test2]; 
SELECT SchemaID, SchemaName, ObjectID, ObjectName, ObjectType, IsMemoryOptimized, IndexID, IndexName, IndexType, AllowPageLocks, IsImageText, IsNewLOB, IsFileStream, IsColumnStore, IsComputed, IsTimestamp, OnReadOnlyFileGroup, ResumableIndexOperation, StatisticsID, StatisticsName, NoRecompute, IsIncremental, PartitionID, PartitionNumber, PartitionCount, [Order], Selected, Completed 
FROM (SELECT schemas.[schema_id] AS SchemaID, schemas.[name] AS SchemaName, objects.[object_id] AS ObjectID, objects.[name] AS ObjectName, RTRIM(objects.[type]) AS ObjectType, tables.is_memory_optimized AS IsMemoryOptimized, indexes.index_id AS IndexID, indexes.[name] AS IndexName, indexes.[type] AS IndexType, indexes.allow_page_locks AS AllowPageLocks, 
CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM sys.columns columns INNER JOIN sys.types types ON columns.system_type_id = types.user_type_id WHERE columns.[object_id] = objects.object_id AND types.name IN('image','text','ntext')) THEN 1 ELSE 0 END AS IsImageText, 
CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM sys.columns columns INNER JOIN sys.types types ON columns.system_type_id = types.user_type_id OR (columns.user_type_id = types.user_type_id AND types.is_Assembly_type = 1) WHERE columns.[object_id] = objects.object_id AND (types.name IN('xml') OR (types.name IN('varchar','nvarchar','varbinary') AND columns.max_length = -1) OR (types.is_Assembly_type = 1 AND columns.max_length = -1))) THEN 1 WHEN indexes.[type] = 2 
            AND EXISTS(SELECT * FROM sys.index_columns index_columns INNER JOIN sys.columns columns ON index_columns.[object_id] = columns.[object_id] 
            AND index_columns.column_id = columns.column_id 
            INNER JOIN sys.types types ON columns.system_type_id = types.user_type_id OR (columns.user_type_id = types.user_type_id AND types.is_Assembly_type = 1) WHERE index_columns.[object_id] = objects.object_id AND index_columns.index_id = indexes.index_id AND (types.[name] IN('xml') OR (types.[name] IN('varchar','nvarchar','varbinary') AND columns.max_length = -1) OR (types.is_Assembly_type = 1 AND columns.max_length = -1))) 
            THEN 1 ELSE 0 END AS IsNewLOB, 
CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM sys.columns columns WHERE columns.[object_id] = objects.object_id  AND columns.is_filestream = 1) THEN 1 ELSE 0 END AS IsFileStream, 
CASE WHEN EXISTS(SELECT * FROM sys.indexes indexes WHERE indexes.[object_id] = objects.object_id AND [type] IN(5,6)) THEN 1 ELSE 0 END AS IsColumnStore, 
CASE WHEN EXISTS(SELECT * FROM sys.index_columns index_columns INNER JOIN sys.columns columns ON index_columns.object_id = columns.object_id AND index_columns.column_id = columns.column_id WHERE (index_columns.key_ordinal > 0 OR index_columns.partition_ordinal > 0) AND columns.is_computed = 1 AND index_columns.object_id = indexes.object_id AND index_columns.index_id = indexes.index_id) THEN 1 ELSE 0 END AS IsComputed, 
CASE WHEN EXISTS(SELECT * FROM sys.index_columns index_columns INNER JOIN sys.columns columns ON index_columns.[object_id] = columns.[object_id] AND index_columns.column_id = columns.column_id INNER JOIN sys.types types ON columns.system_type_id = types.system_type_id 
WHERE index_columns.[object_id] = objects.object_id AND index_columns.index_id = indexes.index_id AND types.[name] = 'timestamp') THEN 1 ELSE 0 END AS IsTimestamp, 
CASE WHEN EXISTS (SELECT * FROM sys.indexes indexes2 INNER JOIN sys.destination_data_spaces destination_data_spaces ON indexes.data_space_id = destination_data_spaces.partition_scheme_id INNER JOIN sys.filegroups filegroups ON destination_data_spaces.data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND indexes2.[object_id] = indexes.[object_id] AND indexes2.[index_id] = indexes.index_id AND destination_data_spaces.destination_id = partitions.partition_number) THEN 1 WHEN EXISTS (SELECT * FROM sys.indexes indexes2 INNER JOIN sys.filegroups filegroups ON indexes.data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND indexes.[object_id] = indexes2.[object_id] AND indexes.[index_id] = indexes2.index_id) THEN 1 WHEN indexes.[type] = 1 AND EXISTS (SELECT * FROM sys.tables tables INNER JOIN sys.filegroups filegroups ON tables.lob_data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND tables.[object_id] = objects.[object_id]) THEN 1 ELSE 0 END AS OnReadOnlyFileGroup, CASE WHEN EXISTS(SELECT * FROM sys.index_resumable_operations index_resumable_operations WHERE state_desc = 'PAUSED' AND index_resumable_operations.object_id = indexes.object_id AND index_resumable_operations.index_id = indexes.index_id AND (index_resumable_operations.partition_number = partitions.partition_number OR index_resumable_operations.partition_number IS NULL)) THEN 1 ELSE 0 END AS ResumableIndexOperation, stats.stats_id AS StatisticsID, stats.name AS StatisticsName, stats.no_recompute AS NoRecompute, stats.is_incremental AS IsIncremental, partitions.partition_id AS PartitionID, partitions.partition_number AS PartitionNumber, IndexPartitions.partition_count AS PartitionCount, 0 AS [Order], 0 AS Selected, 0 AS Completed FROM sys.indexes indexes INNER JOIN sys.objects objects ON indexes.[object_id] = objects.[object_id] INNER JOIN sys.schemas schemas ON objects.[schema_id] = schemas.[schema_id] LEFT OUTER JOIN sys.tables tables ON objects.[object_id] = tables.[object_id] LEFT OUTER JOIN sys.stats stats ON indexes.[object_id] = stats.[object_id] AND indexes.[index_id] = stats.[stats_id] LEFT OUTER JOIN sys.partitions partitions ON indexes.[object_id] = partitions.[object_id] AND indexes.index_id = partitions.index_id LEFT OUTER JOIN (SELECT partitions.[object_id], partitions.index_id, COUNT(DISTINCT partitions.partition_number) AS partition_count FROM sys.partitions partitions GROUP BY partitions.[object_id], partitions.index_id) IndexPartitions ON partitions.[object_id] = IndexPartitions.[object_id] AND partitions.[index_id] = IndexPartitions.[index_id] WHERE objects.[type] IN('U','V') AND objects.is_ms_shipped = 0 AND indexes.[type] IN(1,2,3,4,5,6,7) AND indexes.is_disabled = 0 AND indexes.is_hypothetical = 0 

UNION 

SELECT schemas.[schema_id] AS SchemaID, schemas.[name] AS SchemaName, objects.[object_id] AS ObjectID, objects.[name] AS ObjectName, RTRIM(objects.[type]) AS ObjectType, tables.is_memory_optimized AS IsMemoryOptimized, NULL AS IndexID, NULL AS IndexName, NULL AS IndexType, NULL AS AllowPageLocks, NULL AS IsImageText, NULL AS IsNewLOB, NULL AS IsFileStream, NULL AS IsColumnStore, NULL AS IsComputed, NULL AS IsTimestamp, NULL AS OnReadOnlyFileGroup, NULL AS ResumableIndexOperation, stats.stats_id AS StatisticsID, stats.name AS StatisticsName, stats.no_recompute AS NoRecompute, stats.is_incremental AS IsIncremental, NULL AS PartitionID, dm_db_incremental_stats_properties.partition_number AS PartitionNumber, NULL AS PartitionCount, 0 AS [Order], 0 AS Selected, 0 AS Completed FROM sys.stats stats INNER JOIN sys.objects objects ON stats.[object_id] = objects.[object_id] INNER JOIN sys.schemas schemas ON objects.[schema_id] = schemas.[schema_id] LEFT OUTER JOIN sys.tables tables ON objects.[object_id] = tables.[object_id] OUTER APPLY sys.dm_db_incremental_stats_properties(stats.object_id, stats.stats_id) dm_db_incremental_stats_properties WHERE objects.[type] IN('U','V') AND objects.is_ms_shipped = 0 AND NOT EXISTS(SELECT * FROM sys.indexes indexes WHERE indexes.[object_id] = stats.[object_id] AND indexes.index_id = stats.stats_id)) IndexesStatistics

さて、[dbo].[IndexOptimize]内でこのコマンドをトリガーするものを見つけようとすると、

IF EXISTS(SELECT * FROM @ActionsPreferred) OR @UpdateStatistics IN('ALL','INDEX')

@ActionsPreferredはfragmentationlevelアクションに対応し、@UpdateStatisticsは統計の更新です。つまり、統計の更新やインデックスの再構築/再編成を実行するたびに実行されます。

フィルタリングは適用されず、クエリはデータベース内のすべてのインデックスに対して実行されます。

その結果、現時点では、1つのテーブルを指定したとしても、すべての問題を解決する必要があります。

EXECUTE MNGDB.dbo.IndexOptimize
@Databases = 'test2',
@Indexes =  'test2.dbo.[83631]',
@FragmentationLow = NULL,
@FragmentationMedium = 'INDEX_REORGANIZE,INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationHigh = 'INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationLevel1 = 30,
@FragmentationLevel2 = 50,
@UpdateStatistics = 'ALL',
@OnlyModifiedStatistics = 'Y',
@LogToTable = 'Y',
@LockTimeout = 60,
@EXECUTE= 'N'

enter image description here

hugeコマンドの実行後に実際のインデックスフィルタリングが行われます：

IF @Indexes IS NULL
      BEGIN
        UPDATE tmpIndexesStatistics
        SET tmpIndexesStatistics.Selected = 1
        FROM @tmpIndexesStatistics tmpIndexesStatistics
      END
      ELSE
      BEGIN
...

これらのインデックスに対してインデックス操作を実行するために、それに応じてtmpIndexesStatisticsテーブルを更新します。

このクエリは、私の小さな開発サーバーでそれを強制終了する前に60分以上実行されていました。

これを修正する可能性のあるものは何ですか？

さて、あなたはolaのスクリプトを変更する必要がありますが、一時テーブルを含むソリューションが私のために機能し、1分で実行されました。間違いなく、さらに最適化できます。

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED;
USE [test2]; 

select * 
INTO #Indexes
from sys.indexes;


select *  
INTO #Index_columns
FROM sys.index_columns;

SELECT * 
INTO #columns
FROM SYS.columns;

SELECT * 
INTO #types 
FROM
sys.types;

SELECT *
INTO #destination_data_spaces
FROM sys.destination_data_spaces;


SELECT *
INTO #filegroups
FROM sys.filegroups;


SELECT *
INTO #stats
FROM sys.stats;

SELECT *
INTO #objects
FROM sys.objects;


SELECT *
INTO #partitions
FROM sys.partitions;


SELECT *
INTO #tables
FROM sys.tables;


SELECT *
INTO #index_resumable_operations
FROM sys.index_resumable_operations 

SELECT *
INTO #schemas
FROM sys.schemas 

+ rerun the query with temp tables instead (char limit)

これをインデックス最適化手順で変更するには

申し訳ありませんが、オラはこれを見ると傷つきます

私がプロシージャをスクリプト化したとき、これは1430行目から1537行目までです

 IF DATABASEPROPERTYEX(@CurrentDatabaseName,'Status') = 'ONLINE'
    AND (@CurrentIsDatabaseAccessible = 1 OR @CurrentIsDatabaseAccessible IS NULL)
    AND DATABASEPROPERTYEX(@CurrentDatabaseName,'Updateability') = 'READ_WRITE'
    BEGIN

      -- Select indexes in the current database
      IF (EXISTS(SELECT * FROM @ActionsPreferred) OR @UpdateStatistics IS NOT NULL) AND (GETDATE() < DATEADD(ss,@TimeLimit,@StartTime) OR @TimeLimit IS NULL)
      BEGIN
        SET @CurrentCommand01 = 'SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED;'
                              + 'USE ' + QUOTENAME(@CurrentDatabaseName) + ';
                                select * 
                                INTO #Indexes
                                from sys.indexes;
                                select *  
                                INTO #Index_columns
                                FROM sys.index_columns;
                                SELECT * 
                                INTO #columns
                                FROM SYS.columns;
                                SELECT * 
                                INTO #types 
                                FROM
                                sys.types;
                                SELECT *
                                INTO #destination_data_spaces
                                from sys.destination_data_spaces;
                                SELECT *
                                INTO #filegroups
                                FROM sys.filegroups;
                                SELECT * kill 64
                                INTO #stats
                                from sys.stats;
                                SELECT *
                                INTO #objects
                                from sys.objects;       
                                SELECT *
                                INTO #partitions
                                from sys.partitions;
                                SELECT *
                                INTO #tables
                                FROM sys.tables;
                                SELECT *
                                INTO #index_resumable_operations
                                FROM sys.index_resumable_operations;
                                SELECT *
                                INTO #schemas
                                FROM sys.schemas; '
                              + ' SELECT SchemaID, SchemaName, ObjectID, ObjectName, ObjectType, IsMemoryOptimized, IndexID, IndexName, IndexType, AllowPageLocks, IsImageText, IsNewLOB, IsFileStream, IsColumnStore, IsComputed, IsTimestamp, OnReadOnlyFileGroup, ResumableIndexOperation, StatisticsID, StatisticsName, NoRecompute, IsIncremental, PartitionID, PartitionNumber, PartitionCount, [Order], Selected, Completed'
                              + ' FROM ('

        IF EXISTS(SELECT * FROM @ActionsPreferred) OR @UpdateStatistics IN('ALL','INDEX')
        BEGIN
          SET @CurrentCommand01 = @CurrentCommand01 + 'SELECT schemas.[schema_id] AS SchemaID'
                                                    + ', schemas.[name] AS SchemaName'
                                                    + ', objects.[object_id] AS ObjectID'
                                                    + ', objects.[name] AS ObjectName'
                                                    + ', RTRIM(objects.[type]) AS ObjectType'
                                                    + ', ' + CASE WHEN @Version >= 12 THEN 'tables.is_memory_optimized' ELSE '0' END + ' AS IsMemoryOptimized'
                                                    + ', indexes.index_id AS IndexID'
                                                    + ', indexes.[name] AS IndexName'
                                                    + ', indexes.[type] AS IndexType'
                                                    + ', indexes.allow_page_locks AS AllowPageLocks'

                                                    + ', CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM #columns columns INNER JOIN #types types ON columns.system_type_id = types.user_type_id WHERE columns.[object_id] = objects.object_id AND types.name IN(''image'',''text'',''ntext'')) THEN 1 ELSE 0 END AS IsImageText'

                                                    + ', CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM #columns columns INNER JOIN #types types ON columns.system_type_id = types.user_type_id OR (columns.user_type_id = types.user_type_id AND types.is_Assembly_type = 1) WHERE columns.[object_id] = objects.object_id AND (types.name IN(''xml'') OR (types.name IN(''varchar'',''nvarchar'',''varbinary'') AND columns.max_length = -1) OR (types.is_Assembly_type = 1 AND columns.max_length = -1))) THEN 1'
                                                    + ' WHEN indexes.[type] = 2 AND EXISTS(SELECT * FROM #Index_columns index_columns INNER JOIN #columns columns ON index_columns.[object_id] = columns.[object_id] AND index_columns.column_id = columns.column_id INNER JOIN #types types ON columns.system_type_id = types.user_type_id OR (columns.user_type_id = types.user_type_id AND types.is_Assembly_type = 1) WHERE index_columns.[object_id] = objects.object_id AND index_columns.index_id = indexes.index_id AND (types.[name] IN(''xml'') OR (types.[name] IN(''varchar'',''nvarchar'',''varbinary'') AND columns.max_length = -1) OR (types.is_Assembly_type = 1 AND columns.max_length = -1))) THEN 1 ELSE 0 END AS IsNewLOB'

                                                    + ', CASE WHEN indexes.[type] = 1 AND EXISTS(SELECT * FROM #columns columns WHERE columns.[object_id] = objects.object_id  AND columns.is_filestream = 1) THEN 1 ELSE 0 END AS IsFileStream'

                                                    + ', CASE WHEN EXISTS(SELECT * FROM #Indexes indexes WHERE indexes.[object_id] = objects.object_id AND [type] IN(5,6)) THEN 1 ELSE 0 END AS IsColumnStore'

                                                    + ', CASE WHEN EXISTS(SELECT * FROM #Index_columns index_columns INNER JOIN #columns columns ON index_columns.object_id = columns.object_id AND index_columns.column_id = columns.column_id WHERE (index_columns.key_ordinal > 0 OR index_columns.partition_ordinal > 0) AND columns.is_computed = 1 AND index_columns.object_id = indexes.object_id AND index_columns.index_id = indexes.index_id) THEN 1 ELSE 0 END AS IsComputed'

                                                    + ', CASE WHEN EXISTS(SELECT * FROM #Index_columns index_columns INNER JOIN #columns columns ON index_columns.[object_id] = columns.[object_id] AND index_columns.column_id = columns.column_id INNER JOIN #types types ON columns.system_type_id = types.system_type_id WHERE index_columns.[object_id] = objects.object_id AND index_columns.index_id = indexes.index_id AND types.[name] = ''timestamp'') THEN 1 ELSE 0 END AS IsTimestamp'

                                                    + ', CASE WHEN EXISTS (SELECT * FROM #Indexes indexes2 INNER JOIN #destination_data_spaces destination_data_spaces ON indexes.data_space_id = destination_data_spaces.partition_scheme_id INNER JOIN #filegroups filegroups ON destination_data_spaces.data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND indexes2.[object_id] = indexes.[object_id] AND indexes2.[index_id] = indexes.index_id' + CASE WHEN @PartitionLevel = 'Y' THEN ' AND destination_data_spaces.destination_id = partitions.partition_number' ELSE '' END + ') THEN 1'
                                                    + ' WHEN EXISTS (SELECT * FROM #Indexes indexes2 INNER JOIN #filegroups filegroups ON indexes.data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND indexes.[object_id] = indexes2.[object_id] AND indexes.[index_id] = indexes2.index_id) THEN 1'
                                                    + ' WHEN indexes.[type] = 1 AND EXISTS (SELECT * FROM #tables tables INNER JOIN #filegroups filegroups ON tables.lob_data_space_id = filegroups.data_space_id WHERE filegroups.is_read_only = 1 AND tables.[object_id] = objects.[object_id]) THEN 1 ELSE 0 END AS OnReadOnlyFileGroup'

                                                    + ', ' + CASE WHEN @Version >= 14 THEN 'CASE WHEN EXISTS(SELECT * FROM #index_resumable_operations index_resumable_operations WHERE state_desc = ''PAUSED'' AND index_resumable_operations.object_id = indexes.object_id AND index_resumable_operations.index_id = indexes.index_id AND (index_resumable_operations.partition_number = partitions.partition_number OR index_resumable_operations.partition_number IS NULL)) THEN 1 ELSE 0 END' ELSE '0' END + ' AS ResumableIndexOperation'

                                                    + ', stats.stats_id AS StatisticsID'
                                                    + ', stats.name AS StatisticsName'
                                                    + ', stats.no_recompute AS NoRecompute'
                                                    + ', ' + CASE WHEN @Version >= 12 THEN 'stats.is_incremental' ELSE '0' END + ' AS IsIncremental'
                                                    + ', ' + CASE WHEN @PartitionLevel = 'Y' THEN 'partitions.partition_id AS PartitionID' WHEN @PartitionLevel = 'N' THEN 'NULL AS PartitionID' END
                                                    + ', ' + CASE WHEN @PartitionLevel = 'Y' THEN 'partitions.partition_number AS PartitionNumber' WHEN @PartitionLevel = 'N' THEN 'NULL AS PartitionNumber' END
                                                    + ', ' + CASE WHEN @PartitionLevel = 'Y' THEN 'IndexPartitions.partition_count AS PartitionCount' WHEN @PartitionLevel = 'N' THEN 'NULL AS PartitionCount' END
                                                    + ', 0 AS [Order]'
                                                    + ', 0 AS Selected'
                                                    + ', 0 AS Completed'
                                                    + ' FROM #Indexes indexes'
                                                    + ' INNER JOIN #objects objects ON indexes.[object_id] = objects.[object_id]'
                                                    + ' INNER JOIN #schemas schemas ON objects.[schema_id] = schemas.[schema_id]'
                                                    + ' LEFT OUTER JOIN #tables tables ON objects.[object_id] = tables.[object_id]'
                                                    + ' LEFT OUTER JOIN #stats stats ON indexes.[object_id] = stats.[object_id] AND indexes.[index_id] = stats.[stats_id]'
          IF @PartitionLevel = 'Y'
          BEGIN
            SET @CurrentCommand01 = @CurrentCommand01 + ' LEFT OUTER JOIN #partitions partitions ON indexes.[object_id] = partitions.[object_id] AND indexes.index_id = partitions.index_id'
                                                      + ' LEFT OUTER JOIN (SELECT partitions.[object_id], partitions.index_id, COUNT(DISTINCT partitions.partition_number) AS partition_count FROM #partitions partitions GROUP BY partitions.[object_id], partitions.index_id) IndexPartitions ON partitions.[object_id] = IndexPartitions.[object_id] AND partitions.[index_id] = IndexPartitions.[index_id]'
          END

          SET @CurrentCommand01 = @CurrentCommand01 + ' WHERE objects.[type] IN(''U'',''V'')'
                                                    + CASE WHEN @MSShippedObjects = 'N' THEN ' AND objects.is_ms_shipped = 0' ELSE '' END
                                                    + ' AND indexes.[type] IN(1,2,3,4,5,6,7)'
                                                    + ' AND indexes.is_disabled = 0 AND indexes.is_hypothetical = 0'
        END

1つのテーブルでプロシージャをテストして再実行すると、1分で実行されました。

EXECUTE dbo.IndexOptimize
@Databases = 'test2',
@Indexes =  'test2.dbo.[83631]',
@FragmentationLow = NULL,
@FragmentationMedium = 'INDEX_REORGANIZE,INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationHigh = 'INDEX_REBUILD_ONLINE,INDEX_REBUILD_OFFLINE',
@FragmentationLevel1 = 30,
@FragmentationLevel2 = 50,
@UpdateStatistics = 'ALL',
@OnlyModifiedStatistics = 'Y',
@LogToTable = 'Y',
@LockTimeout = 60,
@EXECUTE= 'N';

期間00:01:18

以前のフィルター/インデックス付き一時テーブル/ ...を使用して、クエリをさらに改善できます。

5
Randi Vertongen

これは、データベースのサイズを知らずに答えることは非常に困難です。最後のメンテナンスジョブがこのサーバーで実行されたとき、インデックスの現在の断片化レベルは何ですか。上記の情報はいずれも提供されていないため、一般的な用語で答えようと思います。

Standardエディションを使用しているため、インデックスをオンラインで再構築するオプションはありません。オフピーク時にインデックス数を含むいくつかのテーブルを（カンマ区切りを使用して）渡すことにより、IndexOptimizeプロシージャを手動で実行してみて、どのように実行されるかを確認できます。上記のコードキャプチャでは、選択されたスキーマはdboです。つまり、dbデータベースで複数のスキーマが定義されていない限り、すべてのオブジェクトが含まれます。また、サイズに基づいてテーブルを選択する際に大きな役割を果たす@pagecountlevelも指定していません。大量のテーブルのみをターゲットにする場合は、この数をかなり高く設定する必要があります。

IndexOptimizeプロシージャのさまざまなパラメータとその重要性について説明する記事を以下に示します。

https://www.sqlshack.com/ola-hallengrens-sql-server-maintenance-solution-index-statistics-maintenance/

ヒープとして多数のテーブルがある場合、Olaのスクリプトはテーブルの再構築を行いません。このスクリプトのさまざまな側面については、以下のBrent Ozarの記事を参照してください。

デフォルトを変更し、これに従ってカスタマイズします article

  • FragmentationLevel1 = 50％

  • FragmentationLevel2 = 80％

  • FragmentationMedium = ‘INDEX_REORGANIZE、INDEX_REBUILD_ONLINE’

  • FragmentationHigh = ‘INDEX_REBUILD_ONLINE’

パフォーマンスの問題の詳細については、以下のリンクも参照してください。

https://www.brentozar.com/archive/2017/12/index-maintenance-madness/

彼のスクリプトを使用すると、インデックスの再編成/再構築を行うよりも頻繁に統計のみを更新できます。これについては、以下で説明します。

https://www.sqlskills.com/blogs/erin/updating-statistics-with-ola-hallengrens-script/

上記で問題が解決した場合はお知らせください。

1
Learning_DBAdmin