SQLSERVER資料庫死鎖與優化雜談

摘要： 死鎖雜談 當資料庫死鎖時，SqlServer會釋放一個優先順序較低的鎖，讓另一個事務執行；所以，即時去捕捉資料庫死鎖，是挺不容易的。 如果，資料庫死鎖比較長時間，那麼死鎖是可以被捕捉的。 可以用SqlServer活動監視器來檢視，哪些程序鎖了資料庫。 首先開啟SqlServer...

死鎖雜談

當資料庫死鎖時，SqlServer會釋放一個優先順序較低的鎖，讓另一個事務執行；所以，即時去捕捉資料庫死鎖，是挺不容易的。

如果，資料庫死鎖比較長時間，那麼死鎖是可以被捕捉的。

可以用SqlServer活動監視器來檢視，哪些程序鎖了資料庫。

首先開啟SqlServer活動監視器，然後可以看到，介面裡有程序，檢視資源，資料檔案I/O，最近消耗大量資源的查詢四項。

四項顯示內容如下：

程序：在程序裡可以看到哪些程序被阻塞，檢視屬性【阻塞者】可以看到，【阻塞者】的會話ID。

等待資源：等待資源裡有一些鎖，可以看看那些鎖累計等待時間較多。

資料檔案I/O：資料檔案I/O記錄一些資料庫MDF，LDF的讀寫速度。

最近消耗大量資源的查詢：記錄一些消耗資源較大的SQL查詢。

查詢程序裡被死鎖的會話ID，然後執行下面的SQL，進行解鎖。

declare @spid int Set @spid = 518 --鎖表程序會話ID 
declare @sql varchar(1000) 
set @sql='kill '+cast(@spid as varchar) 
exec(@sql)

也可以用下面SQL語句查詢死鎖程序，這樣查詢死鎖程序，定位比較快。

selectrequest_session_id spid,OBJECT_NAME(resource_associated_entity_id) tableName
fromsys.dm_tran_locks whereresource_type='OBJECT'

優化雜談

最近消耗大量資源的查詢也可以用SQL查詢。

下面SQL是查詢最耗時的前10條SQL語句。

SELECT TOP 10 total_worker_time / 1000 AS [自編譯以來執行所用的CPU時間總量(ms-毫秒)],
total_elapsed_time/1000 as [完成執行此計劃所用的總時間],
total_elapsed_time / execution_count/1000 as [平均完成執行此計劃所用時間],
execution_countas [上次編譯以來所執行的次數],
creation_time as [編譯計劃的時間],
deqs.total_worker_time / deqs.execution_count / 1000 AS [平均使用CPU時間(ms)],
last_execution_time AS [上次開始執行計劃的時間],
total_physical_reads [編譯後在執行期間所執行的物理讀取總次數],
total_logical_reads/execution_count [平均邏輯讀次數],
min_worker_time /1000 AS [單次執行期間所用的最小CPU時間(ms)],
max_worker_time / 1000 AS [單次執行期間所用的最大 CPU 時間(ms)],
SUBSTRING(dest.text, deqs.statement_start_offset / 2 + 1,
(CASE WHEN deqs.statement_end_offset = -1 THEN DATALENGTH(dest.text) ELSE deqs.statement_end_offset END - deqs.statement_start_offset) / 2 + 1) AS [執行SQL], 
dest.text as [完整SQL],
db_name(dest.dbid) as [資料庫名稱],
object_name(dest.objectid, dest.dbid) as [物件名稱]
,deqs.plan_handle [查詢所屬的已編譯計劃]
FROM sys.dm_exec_query_stats deqs WITH(NOLOCK)
CROSS APPLY sys.dm_exec_sql_text(deqs.sql_handle) AS dest --平均使用CPU時間降序 
ORDER BY (deqs.total_worker_time / deqs.execution_count / 1000) DESC

在SqlServer活動監視器裡，檢視資源等待。

通常可以看到等待類別是Latch的排在最上面，如下圖：

Latch 【閂鎖】雖然是一種輕量級的鎖，但等待的鎖越多，肯定越影響資料庫效能。

執行下面SQL，檢視下哪些Latch比較耗資源。

SELECT * FROM sys.dm_os_latch_stats

查詢結果如下圖所示：

從結果中可以看到各種鎖型別的請求的次數，等待時間，最大等待時間（毫秒）。

但這些鎖型別都是英文簡寫，需要使用下面表格查詢它們的真實意義。

通過對比表格，我們發現了最消耗資源的ACCESS_METHODS_DATASET_PARENT鎖的意義是併發操作時資源訪問的鎖。那麼想降低併發操作，就可以減少ACCESS_METHODS_DATASET_PARENT鎖的資源消耗了。

Latch參考網址： https://docs.microsoft.com/en-us/sql/relational-databases/system-dynamic-management-views/sys-dm-os-latch-stats-transact-sql?view=sql-server-2017

Latch class	Description
ALLOC_CREATE_RINGBUF	Used internally by SQL Server to initialize the synchronization of the creation of an allocation ring buffer.
ALLOC_CREATE_FREESPACE_CACHE	Used to initialize the synchronization of internal freespace caches for heaps.
ALLOC_CACHE_MANAGER	Used to synchronize internal coherency tests.
ALLOC_FREESPACE_CACHE	Used to synchronize the access to a cache of pages with available space for heaps and binary large objects (BLOBs). Contention on latches of this class can occur when multiple connections try to insert rows into a heap or BLOB at the same time. You can reduce this contention by partitioning the object. Each partition has its own latch. Partitioning will distribute the inserts across multiple latches.
ALLOC_EXTENT_CACHE	Used to synchronize the access to a cache of extents that contains pages that are not allocated. Contention on latches of this class can occur when multiple connections try to allocate data pages in the same allocation unit at the same time. This contention can be reduced by partitioning the object of which this allocation unit is a part.
ACCESS_METHODS_DATASET_PARENT	Used to synchronize child dataset access to the parent dataset during parallel operations.
ACCESS_METHODS_HOBT_FACTORY	Used to synchronize access to an internal hash table.
ACCESS_METHODS_HOBT	Used to synchronize access to the in-memory representation of a HoBt.
ACCESS_METHODS_HOBT_COUNT	Used to synchronize access to a HoBt page and row counters.
ACCESS_METHODS_HOBT_VIRTUAL_ROOT	Used to synchronize access to the root page abstraction of an internal B-tree.
ACCESS_METHODS_CACHE_ONLY_HOBT_ALLOC	Used to synchronize worktable access.
ACCESS_METHODS_BULK_ALLOC	Used to synchronize access within bulk allocators.
ACCESS_METHODS_SCAN_RANGE_GENERATOR	Used to synchronize access to a range generator during parallel scans.
ACCESS_METHODS_KEY_RANGE_GENERATOR	Used to synchronize access to read-ahead operations during key range parallel scans.
APPEND_ONLY_STORAGE_INSERT_POINT	Used to synchronize inserts in fast append-only storage units.
APPEND_ONLY_STORAGE_FIRST_ALLOC	Used to synchronize the first allocation for an append-only storage unit.
APPEND_ONLY_STORAGE_UNIT_MANAGER	Used for internal data structure access synchronization within the fast append-only storage unit manager.
APPEND_ONLY_STORAGE_MANAGER	Used to synchronize shrink operations in the fast append-only storage unit manager.
BACKUP_RESULT_SET	Used to synchronize parallel backup result sets.
BACKUP_TAPE_POOL	Used to synchronize backup tape pools.
BACKUP_LOG_REDO	Used to synchronize backup log redo operations.
BACKUP_INSTANCE_ID	Used to synchronize the generation of instance IDs for backup performance monitor counters.
BACKUP_MANAGER	Used to synchronize the internal backup manager.
BACKUP_MANAGER_DIFFERENTIAL	Used to synchronize differential backup operations with DBCC.
BACKUP_OPERATION	Used for internal data structure synchronization within a backup operation, such as database, log, or file backup.
BACKUP_FILE_HANDLE	Used to synchronize file open operations during a restore operation.
BUFFER	Used to synchronize short term access to database pages. A buffer latch is required before reading or modifying any database page. Buffer latch contention can indicate several issues, including hot pages and slow I/Os. This latch class covers all possible uses of page latches. sys.dm_os_wait_stats makes a difference between page latch waits that are caused by I/O operations and read and write operations on the page.
BUFFER_POOL_GROW	Used for internal buffer manager synchronization during buffer pool grow operations.
DATABASE_CHECKPOINT	Used to serialize checkpoints within a database.
CLR_PROCEDURE_HASHTABLE	Internal use only.
CLR_UDX_STORE	Internal use only.
CLR_DATAT_ACCESS	Internal use only.
CLR_XVAR_PROXY_LIST	Internal use only.
DBCC_CHECK_AGGREGATE	Internal use only.
DBCC_CHECK_RESULTSET	Internal use only.
DBCC_CHECK_TABLE	Internal use only.
DBCC_CHECK_TABLE_INIT	Internal use only.
DBCC_CHECK_TRACE_LIST	Internal use only.
DBCC_FILE_CHECK_OBJECT	Internal use only.
DBCC_PERF	Used to synchronize internal performance monitor counters.
DBCC_PFS_STATUS	Internal use only.
DBCC_OBJECT_METADATA	Internal use only.
DBCC_HASH_DLL	Internal use only.
EVENTING_CACHE	Internal use only.
FCB	Used to synchronize access to the file control block.
FCB_REPLICA	Internal use only.
FGCB_ALLOC	Use to synchronize access to round robin allocation information within a filegroup.
FGCB_ADD_REMOVE	Use to synchronize access to filegroups for add, drop, grow, and shrink file operations.
FILEGROUP_MANAGER	Internal use only.
FILE_MANAGER	Internal use only.
FILESTREAM_FCB	Internal use only.
FILESTREAM_FILE_MANAGER	Internal use only.
FILESTREAM_GHOST_FILES	Internal use only.
FILESTREAM_DFS_ROOT	Internal use only.
LOG_MANAGER	Internal use only.
FULLTEXT_DOCUMENT_ID	Internal use only.
FULLTEXT_DOCUMENT_ID_TRANSACTION	Internal use only.
FULLTEXT_DOCUMENT_ID_NOTIFY	Internal use only.
FULLTEXT_LOGS	Internal use only.
FULLTEXT_CRAWL_LOG	Internal use only.
FULLTEXT_ADMIN	Internal use only.
FULLTEXT_AMDIN_COMMAND_CACHE	Internal use only.
FULLTEXT_LANGUAGE_TABLE	Internal use only.
FULLTEXT_CRAWL_DM_LIST	Internal use only.
FULLTEXT_CRAWL_CATALOG	Internal use only.
FULLTEXT_FILE_MANAGER	Internal use only.
DATABASE_MIRRORING_REDO	Internal use only.
DATABASE_MIRRORING_SERVER	Internal use only.
DATABASE_MIRRORING_CONNECTION	Internal use only.
DATABASE_MIRRORING_STREAM	Internal use only.
QUERY_OPTIMIZER_VD_MANAGER	Internal use only.
QUERY_OPTIMIZER_ID_MANAGER	Internal use only.
QUERY_OPTIMIZER_VIEW_REP	Internal use only.
RECOVERY_BAD_PAGE_TABLE	Internal use only.
RECOVERY_MANAGER	Internal use only.
SECURITY_OPERATION_RULE_TABLE	Internal use only.
SECURITY_OBJPERM_CACHE	Internal use only.
SECURITY_CRYPTO	Internal use only.
SECURITY_KEY_RING	Internal use only.
SECURITY_KEY_LIST	Internal use only.
SERVICE_BROKER_CONNECTION_RECEIVE	Internal use only.
SERVICE_BROKER_TRANSMISSION	Internal use only.
SERVICE_BROKER_TRANSMISSION_UPDATE	Internal use only.
SERVICE_BROKER_TRANSMISSION_STATE	Internal use only.
SERVICE_BROKER_TRANSMISSION_ERRORS	Internal use only.
SSBXmitWork	Internal use only.
SERVICE_BROKER_MESSAGE_TRANSMISSION	Internal use only.
SERVICE_BROKER_MAP_MANAGER	Internal use only.
SERVICE_BROKER_HOST_NAME	Internal use only.
SERVICE_BROKER_READ_CACHE	Internal use only.
SERVICE_BROKER_WAITFOR_MANAGER	Used to synchronize an instance level map of waiter queues. One queue exists per database ID, Database Version, and Queue ID tuple. Contention on latches of this class can occur when many connections are: In a WAITFOR(RECEIVE) wait state; calling WAITFOR(RECEIVE); exceeding the WAITFOR timeout; receiving a message; committing or rolling back the transaction that contains the WAITFOR(RECEIVE); You can reduce the contention by reducing the number of threads in a WAITFOR(RECEIVE) wait state.
SERVICE_BROKER_WAITFOR_TRANSACTION_DATA	Internal use only.
SERVICE_BROKER_TRANSMISSION_TRANSACTION_DATA	Internal use only.
SERVICE_BROKER_TRANSPORT	Internal use only.
SERVICE_BROKER_MIRROR_ROUTE	Internal use only.
TRACE_ID	Internal use only.
TRACE_AUDIT_ID	Internal use only.
TRACE	Internal use only.
TRACE_CONTROLLER	Internal use only.
TRACE_EVENT_QUEUE	Internal use only.
TRANSACTION_DISTRIBUTED_MARK	Internal use only.
TRANSACTION_OUTCOME	Internal use only.
NESTING_TRANSACTION_READONLY	Internal use only.
NESTING_TRANSACTION_FULL	Internal use only.
MSQL_TRANSACTION_MANAGER	Internal use only.
DATABASE_AUTONAME_MANAGER	Internal use only.
UTILITY_DYNAMIC_VECTOR	Internal use only.
UTILITY_SPARSE_BITMAP	Internal use only.
UTILITY_DATABASE_DROP	Internal use only.
UTILITY_DYNAMIC_MANAGER_VIEW	Internal use only.
UTILITY_DEBUG_FILESTREAM	Internal use only.
UTILITY_LOCK_INFORMATION	Internal use only.
VERSIONING_TRANSACTION	Internal use only.
VERSIONING_TRANSACTION_LIST	Internal use only.
VERSIONING_TRANSACTION_CHAIN	Internal use only.
VERSIONING_STATE	Internal use only.
VERSIONING_STATE_CHANGE	Internal use only.
KTM_VIRTUAL_CLOCK	Internal use only.

DBCC雜談

DBCC 語句是 SQL Server 的資料庫控制檯命令，共有以下四種類型。

維護：對資料庫、索引或檔案組進行維護的任務。

雜項：雜項任務，如啟用跟蹤標誌或從記憶體中刪除 DLL。

資訊：收集並顯示各種型別資訊的任務。

驗證：對資料庫、表、索引、目錄、檔案組或資料庫頁的分配進行的驗證操作。

DBCC shrinkdatabase

DBCC shrinkdatabase用於收縮資料庫，SQL語句如下：

DBCC shrinkdatabase (N'庫名' , 1)

執行結果如下：

各欄位含義如下：

DbId：資料庫引擎試圖收縮的檔案的資料庫標識號。

FileId：資料庫引擎嘗試收縮的檔案的檔案標識號。

CurrentSize：檔案當前佔用的 8 KB 頁數。

MinimumSize：檔案最低可以佔用的 8 KB 頁數。 這與檔案的最小大小或最初建立時的大小相對應。

UsedPages：檔案當前使用的 8 KB 頁數。

EstimatedPages：資料庫引擎估計檔案能夠收縮到的 8 KB 頁數。

如果收縮不成功，可以檢視下資料庫是否有可以收縮的空間。

SQL如下：

SELECT name ,size/128.0 - CAST(FILEPROPERTY(name, 'SpaceUsed') AS int)/128.0 AS AvailableSpaceInMB
FROM sys.database_files;

如果有空間還收縮不成功，則可能是別原因。

DBCC參考網址： https://docs.microsoft.com/zh-cn/sql/t-sql/database-console-commands/dbcc-shrinkdatabase-transact-sql?view=sql-server-2017

資料庫日誌雜談

SqlServer資料庫日誌對執行的SQL語句進行了加密，所以，在日誌裡，我們看不到真正執行的SQL語句。

如果想檢視SQL語句，需要藉助一些工具，如ApexSQLLog。

不過，雖然看不到SQL語句，也可以通過日誌看出一些資料庫問題，比如，可以檢視資料庫執行了多少次插入，更新等操作。

檢視資料庫日誌的SQL如下：

SELECT * FROM [sys].[fn_dblog](NULL,NULL)

查詢結果如下：

查詢結果各欄位含義如下：

Operation	Context	解釋
LOP_SET_BITS	LCX_DIFF_MAP	設定點陣圖，資料： 差異(Differential)備份:只備份上次完整備份後，做修改的部分。備份單位是區(Extent)。意味著某個區內即使只有一頁做了變動，則在差異備份裡會被體現.差異備份依靠一個BitMap進行維護，一個Bit對應一個區，自上次完整備份後，被修改的區會被置為1，而BitMap中被置為1對應的區會被差異備份所備份。而到下一次完整備份後，BitMap中所有的Bit都會被重置為0 而這個BitMap在資料庫第7頁： DCM頁 差異變更(Differential Changed Map，DCM)頁面他跟蹤一個檔案中的哪一個區在最新一次完整資料庫備份之後被修改過。SQLSERVER用在增量備份時只對已發生資料變更的分割槽進行增量備份即可
LOP_BEGIN_XACT		事務開始
LOP_MODIFY_ROW	LCX_HEAP	修改堆表中的某一行記錄
LOP_PREP_XACT		準備啟動資料庫
LOP_COMMIT_XACT		提交事務
LOP_MODIFY_ROW	LCX_BOOT_PAGE	修改資料庫啟動頁
LOP_MODIFY_HEADER	LCX_PFS	修改PFS頁的頁頭部資訊
LOP_INSERT_ROWS	LCX_CLUSTERED	插入資料到聚集索引的索引頁
LOP_INSERT_ROWS	LCX_INDEX_LEAF	插入資料到索引的葉子節點即資料頁
LOP_FORMAT_PAGE	LCX_CLUSTERED	重新組織聚集索引
LOP_DELETE_SPLIT	LCX_CLUSTERED	刪除聚集索引表的一行記錄引起頁拆分
LOP_MODIFY_HEADER	LCX_HEAP	修改堆表的某頁的頁頭資訊
LOP_BEGIN_CKPT	LCX_NULL	檢查點開始
LOP_END_CKPT	LCX_NULL	檢查點結束
LOP_SET_FREE_SPACE	LCX_PFS	修改PFS頁設定那個資料頁是空閒的
LOP_ROOT_CHANGE	LCX_CLUSTERED	聚集索引的根節點改變
LOP_INSERT_ROWS	LCX_HEAP	插入資料到堆表
LOP_FORMAT_PAGE	LCX_HEAP	格式化堆裡的資料頁
LOP_LOCK_XACT		在事務裡獲取鎖
LOP_FORMAT_PAGE	LCX_HEAP	格式化堆裡的資料頁