在 Azure Data Studio 和 SSMS 或 SSRS 中执行 SQL 查询时的不同行为

Question

我有一个查询，其行为因执行它的客户端而异。我使用 Azure Data Studio 获得了即时结果（163 行）。 SQL Management Studio、Report Builder、SSRS Web Portal 超过 7 分钟。

更好的是，如果我从 select 中删除两列或添加 2 个空列，我在 SSMS 中实现了相同的性能。

SELECT
  mycols,
  (...),
  null,
  null
FROM
(...)

解释计划一样，我一个都看不懂。调优顾问没有任何建议。

我认为这与客户选择的原因有关，唯一不同的是 ARITHMETIC_ABORT_ENABLED 是假的。在 SSMS 中设置为 true 后仍然没有效果。

我使用 SQL Server 2019。

这里是查询：

DECLARE @UserID as VARCHAR(30) = CURRENT_USER 
DECLARE @ProductionCenterSwipCode as VARCHAR(30) = 'AIX'
DECLARE @ResourceName as VARCHAR(30) = NULL
DECLARE @ActivityCode as VARCHAR(30) = '{ALL}'
DECLARE @SubactivityCode as VARCHAR(30) = '{ALL}'
DECLARE @WorkNatureCode as VARCHAR(30) = '{ALL}'
DECLARE @ClientId as INT = -9999
DECLARE @ContractId as INT = -9999
DECLARE @StartDate as DATE= CAST( '2021-04-01' AS DATE)
DECLARE @EndDate as DATE = CAST( '2021-05-30' AS DATE)


SELECT concat(cct.cct_activity_domain_code, ' - ', cct.cct_activity_domain_label) domain_of_activity,
       concat(hrh.hrh_production_center_swip_code, ' - ', hrh.hrh_production_center_name) resource_production_center,
       hrh.hrh_full_name resource,
       dat.dat_date Date_logged,
       tim.tim_number_hours_logged time_logged,
       concat(rng.rng_type_of_work_code, ' - ', rng.rng_type_of_work_label) type_of_work,
       tsk.tsk_code task,
       concat(sta.wat_activity_code , ' - ', sta.wat_activity_label) activity,
       cct.cct_client_reference client_contract,
       cli.cli_name client_name,
       wko.wko_swip_id wo_id,
       wko.wko_client_reference wo_client_reference,
       concat(sta.wat_workorder_type_code, ' - ', sta.wat_workorder_type_label) wo_type,
       wko.wko_current_status_label wo_status,
       rng.rng_label range_element,
       rnh.rnh_current_status_label range_element_status,
       concat(sta.wat_subactivity_code, ' - ', sta.wat_subactivity_label) sub_activity,
       concat(sta.wat_nature_of_work_code, ' - ', sta.wat_nature_of_work_label) work_nature,
       concat(sta.wat_priority_code, ' - ', sta.wat_priority_label) priority,
       concat(sta.wat_complexity_code, ' - ', sta.wat_complexity_label) complexity,
       concat(sta.wat_skill_level_code, ' - ', sta.wat_skill_level_label) skill_level,
       concat(sta.wat_program_code, ' - ', sta.wat_program_label) program,
       concat(sta.wat_perimeter_code, ' - ', sta.wat_perimeter_label) technical_scope,
       concat(sta.wat_customer_production_center_code, ' - ', sta.wat_customer_production_center_label) client_production_center,
       concat(sta.wat_scenario_code, ' - ', sta.wat_scenario_label) production_scenario,
       concat(sta.wat_application_classification_code, ' - ', sta.wat_application_classification_label) application_ranking,
       concat(sta.wat_customer_technical_leader_code, ' - ', sta.wat_customer_technical_leader_label) client_technical_leader,
       wko.wko_current_start_date wo_start_date,
       wko.wko_current_commit_end_date wo_end_date_commitment,
       wko.wko_last_delivery_date wo_last_delivery_date,
       wko.wko_highest_id highest_level_wo_id,
       wko.wko_highest_client_reference highest_level_wo_client_refence,
       wko.wko_batch_reference wo_batch_reference,
       wko.wko_label wo_descriprion
FROM dwh_swip.swip.f_logged_time tim
INNER JOIN dwh_swip.swip.d_human_resource_hstr hrh ON hrh.hrh_id = tim.tim_hrh_id
INNER JOIN dwh_swip.swip.d_range_element rng ON rng.rng_id = tim.tim_rng_id
INNER JOIN dwh_swip.swip.d_date dat ON dat.dat_id = tim.tim_dat_id
INNER JOIN dwh_swip.swip.d_range_element_hstr rnh ON rnh.rnh_id = tim.tim_rnh_id
INNER JOIN dwh_swip.swip.d_task tsk ON tsk.tsk_id = tim.tim_tsk_id
INNER JOIN dwh_swip.swip.d_workorder wko ON wko.wko_id = tim.tim_wko_id
INNER JOIN dwh_swip.swip.d_client_contract cct ON cct.cct_id = tim.tim_cct_id
INNER JOIN dwh_swip.swip.d_client cli ON cli.cli_id = tim.tim_cli_id
INNER JOIN dwh_swip.swip.d_wo_structuring_attributes sta ON sta.wat_id = tim.tim_wat_id
WHERE exists ( SELECT 1
    FROM swip.b_user_perimeter
    WHERE upr_domain_id = (@UserID) AND upr_active_domain_code = cct.cct_activity_domain_code )
AND hrh.hrh_production_center_swip_code IN (@ProductionCenterSwipCode)
AND cct.cct_activity_domain_code IN (N'ADS_TD',N'ADS_VV_ILL',N'AC_TD',N'BI',N'DEF_TD',N'DOM_MG',N'HE_TD',N'INDUSTRY',N'LLA DOM',N'NAV_CU_SUP',N'NAV_ENG',N'NAV_TD',N'non applicable',N'llaxx',N'unknown')
AND (@ResourceName IS NULL or @ResourceName IS NOT NULL AND hrh.hrh_full_name like N'%'+ @ResourceName + N'%')
AND ('{ALL}' in (@ActivityCode) or '{ALL}' NOT IN (@ActivityCode) AND sta.wat_activity_code IN (@ActivityCode))
AND ('{ALL}' in (@SubactivityCode) or '{ALL}' NOT IN (@SubactivityCode) AND sta.wat_subactivity_code IN (@SubactivityCode))
AND ('{ALL}' in (@WorkNatureCode) or '{ALL}' NOT IN (@WorkNatureCode) AND sta.wat_nature_of_work_code IN (@WorkNatureCode))
AND (-9999 in (@ClientId) or -9999 NOT IN (@ClientId) AND tim.tim_cli_id IN (@ClientId))
AND (-9999 in (@ContractId) or -9999 NOT IN (@ContractId) AND tim.tim_cct_id IN (@ContractId))
AND  tim.tim_dat_id >= (YEAR(@StartDate)*10000 + MONTH(@StartDate)*100 + DAY(@StartDate))
AND  tim.tim_dat_id <= (YEAR(@EndDate)*10000 + MONTH(@EndDate)*100 + DAY(@EndDate) )

Answer 1

这是一个 SQL 服务器错误。微软支持的回答：

ISSUE: You were having performance issues with a specific query. On SSMS and SSRS, it was taking longer to finish. On Azure Data Studio and SQL CMD, it was running faster

Fixes a performance issue when a query outputs NULL values following a semi join in SQL Server 2019 for a database with UTF8.

To solve the issue, you have to apply the latest CU on SQL, which is the CU11

起初 Microsoft 很难搞清楚，因为他们当时不知道行为会根据 SQL 客户端而改变。现在他们知道了^^

在 Azure Data Studio 和 SSMS 或 SSRS 中执行 SQL 查询时的不同行为

Different behavior when executing an SQL query in Azure Data Studio and SSMS or SSRS

sql-server

ssms

reporting-services

azure-data-studio