Identifying causes for deadlock and ways to prevent it - azure-sql-database

Question

We have an ELT process wherein we transfer data from Azure Data Factory into a staging table present in Azure SQL database and then trigger a stored procedure that alters schema from staging to the final version.
The above process executes in parallel for multiple entities (at the same time).
But sometimes the job fails intermittently with failure message as deadlock.
Note: we have enabled Azure Log Analytics for Azure SQL database logs.
I have these questions:
Are there any metadata queries like sys.logs to determine the cause of deadlock and jobs getting executed at that instance.
How to identify the exact time and the cause of deadlock leveraging Azure log analytics (KQL queries)
ALTER SCHEMA uses a schema level lock. So, my understanding is the deadlock might be due to this reason but is there a sure shot way to confirm this is the cause for deadlock as the failures are intermittent and not always.
Is there any sys. queries to identify whether the schema is in lock state or not, so we can check the state of schema before altering it in case of parallel executions assuming #3 is the cause.
Is there any correlation between deadlock occurrence and database tier? Because we observed that after increasing the DB tier, the intermittent failure frequency reduced.

Answer 1

After execution of a workload on the server, issue the following statements in Query Editor to find queries still holding locks.
--
-- The pair matching targets report current unpaired events using
-- the sys.dm_xe_session_targets dynamic management view (DMV)
-- in XML format.
-- The following query retrieves the data from the DMV and stores
-- key data in a temporary table to speed subsequent access and
-- retrieval.
--
SELECT
objlocks.value('(action[#name="session_id"]/value)[1]', 'int')
AS session_id,
objlocks.value('(data[#name="database_id"]/value)[1]', 'int')
AS database_id,
objlocks.value('(data[#name="resource_type"]/text)[1]', 'nvarchar(50)' )
AS resource_type,
objlocks.value('(data[#name="resource_0"]/value)[1]', 'bigint')
AS resource_0,
objlocks.value('(data[#name="resource_1"]/value)[1]', 'bigint')
AS resource_1,
objlocks.value('(data[#name="resource_2"]/value)[1]', 'bigint')
AS resource_2,
objlocks.value('(data[#name="mode"]/text)[1]', 'nvarchar(50)')
AS mode,
objlocks.value('(action[#name="sql_text"]/value)[1]', 'varchar(MAX)')
AS sql_text,
CAST(objlocks.value('(action[#name="plan_handle"]/value)[1]', 'varchar(MAX)') AS xml)
AS plan_handle,
CAST(objlocks.value('(action[#name="tsql_stack"]/value)[1]', 'varchar(MAX)') AS xml)
AS tsql_stack
INTO #unmatched_locks
FROM (
SELECT CAST(xest.target_data as xml)
lockinfo
FROM sys.dm_xe_session_targets xest
JOIN sys.dm_xe_sessions xes ON xes.address = xest.event_session_address
WHERE xest.target_name = 'pair_matching' AND xes.name = 'FindBlockers'
) heldlocks
CROSS APPLY lockinfo.nodes('//event[#name="lock_acquired"]') AS T(objlocks)
--
- Join the data acquired from the pairing target with other
- DMVs to return provide additional information about blockers
SELECT ul.*
FROM #unmatched_locks ul
INNER JOIN sys.dm_tran_locks tl ON ul.database_id = tl.resource_database_id AND
ul.resource_type = tl.resource_type
WHERE resource_0 IS NOT NULL
AND session_id IN
(SELECT blocking_session_id FROM sys.dm_exec_requests WHERE
blocking_session_id != 0)
AND tl.request_status='wait'
AND REPLACE(ul.mode, 'LCK_M_', '' ) = tl.request_mode
https://learn.microsoft.com/en-us/sql/relational-databases/extended-events/determine-which-queries-are-holding-locks?view=sql-server-ver16