ALTER WORKLOAD GROUP (Transact-SQL)

Select a product

In the following row, select the product name you're interested in, and only that product's information is displayed.

* SQL Server *  

 

SQL Server and SQL Managed Instance

Changes an existing Resource Governor workload group configuration, and optionally assigns it to a Resource Governor resource pool.

Note

For Azure SQL Managed Instance, you must be in the context of the master database to alter Resource Governor configuration.

Transact-SQL syntax conventions.

Syntax

ALTER WORKLOAD GROUP { group_name | "default" }
[ WITH
    ([ IMPORTANCE = { LOW | MEDIUM | HIGH } ]
      [ [ , ] REQUEST_MAX_MEMORY_GRANT_PERCENT = value ]
      [ [ , ] REQUEST_MAX_CPU_TIME_SEC = value ]
      [ [ , ] REQUEST_MEMORY_GRANT_TIMEOUT_SEC = value ]
      [ [ , ] MAX_DOP = value ]
      [ [ , ] GROUP_MAX_REQUESTS = value ] )
]
[ USING { pool_name | "default" } ]
[ ; ]

Arguments

group_name | "default"

The name of an existing user-defined workload group or the Resource Governor default workload group. Resource Governor creates the "default" and internal groups when SQL Server is installed.

The option "default" must be enclosed by quotation marks ("") or brackets ([]) when used with ALTER WORKLOAD GROUP to avoid conflict with DEFAULT, which is a system reserved word. For more information, see Database Identifiers.

Predefined workload groups and resource pools all use lowercase names, such as "default". This should be taken into account for servers that use case-sensitive collation. Servers with case-insensitive collation, such as SQL_Latin1_General_CP1_CI_AS, will treat "default" and "Default" as the same.

IMPORTANCE = { LOW | MEDIUM | HIGH }

Specifies the relative importance of a request in the workload group. Importance is one of the following:

  • LOW
  • MEDIUM (default)
  • HIGH

Internally, each importance setting is stored as a number that is used for calculations.

IMPORTANCE is local to the resource pool; workload groups of different importance inside the same resource pool affect each other, but don't affect workload groups in another resource pool.

REQUEST_MAX_MEMORY_GRANT_PERCENT = value

Specifies the maximum amount of memory that a single request can take from the pool. value is a percentage relative to the resource pool size specified by MAX_MEMORY_PERCENT. The default value is 25. The amount specified only refers to query execution grant memory.

value is an int up to SQL Server 2017 (14.x) and the allowed range is from 1 through 100. Starting with SQL Server 2019 (15.x), the value is a float data type and the allowed range is from 0 through 100.

Important

Setting value to 0 prevents queries with SORT and HASH JOIN operations in user-defined workload groups from running.

It is not recommended to set value greater than 70 because the server may be unable to set aside enough free memory if other concurrent queries are running. This may eventually lead to query time-out error 8645.

If the query memory requirements exceed the limit that is specified by this parameter, the server does the following:

  • For user-defined workload groups, the server tries to reduce the query degree of parallelism until the memory requirement falls under the limit, or until the degree of parallelism equals 1. If the query memory requirement is still greater than the limit, error 8657 occurs.
  • For internal and default workload groups, the server permits the query to obtain the required memory.

Both cases are subject to time-out error 8645 if the server has insufficient physical memory.

REQUEST_MAX_CPU_TIME_SEC = value

Specifies the maximum amount of CPU time, in seconds, that a request can use. value must be 0 or a positive integer. The default setting for value is 0, which means unlimited. By default, Resource Governor won't prevent a request from continuing if the maximum time is exceeded. However, an event will be generated. For more information, see CPU Threshold Exceeded Event Class.

Starting with SQL Server 2016 (13.x) SP2 and SQL Server 2017 (14.x) CU3, and using trace flag 2422, Resource Governor will abort a request when the maximum time is exceeded.

REQUEST_MEMORY_GRANT_TIMEOUT_SEC = value

Specifies the maximum time, in seconds, that a query can wait for memory grant (work buffer memory) to become available.

A query doesn't always fail when memory grant time-out is reached. A query will only fail if there are too many concurrent queries running. Otherwise, the query may only get the minimum memory grant, resulting in reduced query performance.

value must be a positive integer. The default setting for value, 0, uses an internal calculation based on query cost to determine the maximum time.

MAX_DOP = value

Specifies the maximum degree of parallelism (DOP) for parallel requests. value must be 0 or a positive integer, 1 though 255. When value is 0, the server chooses the max degree of parallelism. This is the default and recommended setting.

The actual value that the Database Engine sets for MAX_DOP by might be less than the specified value. The final value is determined by the formula min(255, number of CPUs).

Caution

Changing MAX_DOP can adversely affect a server's performance. If you must change MAX_DOP, we recommend that it be set to a value that is less than or equal to the maximum number of hardware schedulers that are present in a single NUMA node. We recommend that you do not set MAX_DOP to a value greater than 8.

MAX_DOP is handled as follows:

  • MAX_DOP as a query hint is honored as long as it doesn't exceed workload group MAX_DOP.

  • MAX_DOP as a query hint always overrides sp_configure 'max degree of parallelism'.

  • Workload group MAX_DOP overrides sp_configure 'max degree of parallelism'.

  • If the query is marked as serial (MAX_DOP = 1) at compile time, it can't be changed back to parallel at run time regardless of the workload group or sp_configure setting.

After DOP is configured, it can only be lowered on grant memory pressure. Workload group reconfiguration isn't visible while waiting in the grant memory queue.

GROUP_MAX_REQUESTS = value

Specifies the maximum number of simultaneous requests that are allowed to execute in the workload group. value must be 0 or a positive integer. The default setting for value, 0, allows unlimited requests. When the maximum concurrent requests are reached, a user in that group can sign in, but is placed in a wait state until concurrent requests are dropped below the value specified.

USING { pool_name | "default" }

Associates the workload group with the user-defined resource pool identified by pool_name, which in effect puts the workload group in the resource pool. If pool_name isn't provided or if the USING argument isn't used, the workload group is put in the predefined Resource Governor default pool.

The option "default" is case-sensitive and must be enclosed by quotation marks ("") or brackets ([]) when used with ALTER WORKLOAD GROUP to avoid conflict with DEFAULT, which is a system reserved word. For more information, see Database Identifiers.

Remarks

ALTER WORKLOAD GROUP is allowed on the default group.

Changes to the workload group configuration don't take effect until after ALTER RESOURCE GOVERNOR RECONFIGURE is executed. When changing a plan affecting setting, the new setting will only take effect in previously cached plans after executing DBCC FREEPROCCACHE (*pool_name*), where pool_name is the name of a Resource Governor resource pool on which the workload group is associated with.

  • If changing MAX_DOP to 1, executing DBCC FREEPROCCACHE isn't required because parallel plans can run in serial mode. However, it may not be as efficient as a plan compiled as a serial plan.

  • If changing MAX_DOP from 1 to 0 or a value greater than 1, executing DBCC FREEPROCCACHE isn't required. However, serial plans can't run in parallel, so clearing the respective cache will allow new plans to potentially be compiled using parallelism.

Caution

Clearing cached plans from a resource pool that is associated with more than one workload group will affect all workload groups with the user-defined resource pool identified by pool_name.

When executing DDL statements, you should be familiar with Resource Governor states. For more information, see Resource Governor.

REQUEST_MEMORY_GRANT_PERCENT: In SQL Server 2005 (9.x), index creation is allowed to use more workspace memory than initially granted for improved performance. This special handling is supported by Resource Governor in later versions, however, the initial grant and any additional memory grant are limited by resource pool and workload group settings.

Index creation on a partitioned table

The memory consumed by index creation on non-aligned partitioned table is proportional to the number of partitions involved. If the total required memory exceeds the per-query limit (REQUEST_MAX_MEMORY_GRANT_PERCENT) imposed by the Resource Governor workload group setting, this index creation may fail to execute. Because the "default" workload group allows a query to exceed the per-query limit with the minimum required memory to start for SQL Server 2005 (9.x) compatibility, the user may be able to run the same index creation in "default" workload group, if the "default" resource pool has enough total memory configured to run such query.

Permissions

Requires CONTROL SERVER permission.

Examples

The following example shows how to change the importance of requests in the default group from MEDIUM to LOW.

ALTER WORKLOAD GROUP "default"
WITH (IMPORTANCE = LOW);
GO
ALTER RESOURCE GOVERNOR RECONFIGURE;
GO

The following example shows how to move a workload group from the pool that it's in, to the default pool.

ALTER WORKLOAD GROUP adHoc
USING [default];
GO
ALTER RESOURCE GOVERNOR RECONFIGURE;
GO

See also

* SQL Managed Instance *  

 

SQL Server and SQL Managed Instance

Changes an existing Resource Governor workload group configuration, and optionally assigns it to a Resource Governor resource pool.

Note

For Azure SQL Managed Instance, you must be in the context of the master database to alter Resource Governor configuration.

Transact-SQL syntax conventions.

Syntax

ALTER WORKLOAD GROUP { group_name | "default" }
[ WITH
    ([ IMPORTANCE = { LOW | MEDIUM | HIGH } ]
      [ [ , ] REQUEST_MAX_MEMORY_GRANT_PERCENT = value ]
      [ [ , ] REQUEST_MAX_CPU_TIME_SEC = value ]
      [ [ , ] REQUEST_MEMORY_GRANT_TIMEOUT_SEC = value ]
      [ [ , ] MAX_DOP = value ]
      [ [ , ] GROUP_MAX_REQUESTS = value ] )
]
[ USING { pool_name | "default" } ]
[ ; ]

Arguments

group_name | "default"

The name of an existing user-defined workload group or the Resource Governor default workload group. Resource Governor creates the "default" and internal groups when SQL Server is installed.

The option "default" must be enclosed by quotation marks ("") or brackets ([]) when used with ALTER WORKLOAD GROUP to avoid conflict with DEFAULT, which is a system reserved word. For more information, see Database Identifiers.

Predefined workload groups and resource pools all use lowercase names, such as "default". This should be taken into account for servers that use case-sensitive collation. Servers with case-insensitive collation, such as SQL_Latin1_General_CP1_CI_AS, will treat "default" and "Default" as the same.

IMPORTANCE = { LOW | MEDIUM | HIGH }

Specifies the relative importance of a request in the workload group. Importance is one of the following:

  • LOW
  • MEDIUM (default)
  • HIGH

Internally, each importance setting is stored as a number that is used for calculations.

IMPORTANCE is local to the resource pool; workload groups of different importance inside the same resource pool affect each other, but don't affect workload groups in another resource pool.

REQUEST_MAX_MEMORY_GRANT_PERCENT = value

Specifies the maximum amount of memory that a single request can take from the pool. value is a percentage relative to the resource pool size specified by MAX_MEMORY_PERCENT. The default value is 25. The amount specified only refers to query execution grant memory.

value is an int up to SQL Server 2017 (14.x) and the allowed range is from 1 through 100. Starting with SQL Server 2019 (15.x), the value is a float data type and the allowed range is from 0 through 100.

Important

Setting value to 0 prevents queries with SORT and HASH JOIN operations in user-defined workload groups from running.

It is not recommended to set value greater than 70 because the server may be unable to set aside enough free memory if other concurrent queries are running. This may eventually lead to query time-out error 8645.

If the query memory requirements exceed the limit that is specified by this parameter, the server does the following:

  • For user-defined workload groups, the server tries to reduce the query degree of parallelism until the memory requirement falls under the limit, or until the degree of parallelism equals 1. If the query memory requirement is still greater than the limit, error 8657 occurs.
  • For internal and default workload groups, the server permits the query to obtain the required memory.

Both cases are subject to time-out error 8645 if the server has insufficient physical memory.

REQUEST_MAX_CPU_TIME_SEC = value

Specifies the maximum amount of CPU time, in seconds, that a request can use. value must be 0 or a positive integer. The default setting for value is 0, which means unlimited. By default, Resource Governor won't prevent a request from continuing if the maximum time is exceeded. However, an event will be generated. For more information, see CPU Threshold Exceeded Event Class.

Starting with SQL Server 2016 (13.x) SP2 and SQL Server 2017 (14.x) CU3, and using trace flag 2422, Resource Governor will abort a request when the maximum time is exceeded.

REQUEST_MEMORY_GRANT_TIMEOUT_SEC = value

Specifies the maximum time, in seconds, that a query can wait for memory grant (work buffer memory) to become available.

A query doesn't always fail when memory grant time-out is reached. A query will only fail if there are too many concurrent queries running. Otherwise, the query may only get the minimum memory grant, resulting in reduced query performance.

value must be a positive integer. The default setting for value, 0, uses an internal calculation based on query cost to determine the maximum time.

MAX_DOP = value

Specifies the maximum degree of parallelism (DOP) for parallel requests. value must be 0 or a positive integer, 1 though 255. When value is 0, the server chooses the max degree of parallelism. This is the default and recommended setting.

The actual value that the Database Engine sets for MAX_DOP by might be less than the specified value. The final value is determined by the formula min(255, number of CPUs).

Caution

Changing MAX_DOP can adversely affect a server's performance. If you must change MAX_DOP, we recommend that it be set to a value that is less than or equal to the maximum number of hardware schedulers that are present in a single NUMA node. We recommend that you do not set MAX_DOP to a value greater than 8.

MAX_DOP is handled as follows:

  • MAX_DOP as a query hint is honored as long as it doesn't exceed workload group MAX_DOP.

  • MAX_DOP as a query hint always overrides sp_configure 'max degree of parallelism'.

  • Workload group MAX_DOP overrides sp_configure 'max degree of parallelism'.

  • If the query is marked as serial (MAX_DOP = 1) at compile time, it can't be changed back to parallel at run time regardless of the workload group or sp_configure setting.

After DOP is configured, it can only be lowered on grant memory pressure. Workload group reconfiguration isn't visible while waiting in the grant memory queue.

GROUP_MAX_REQUESTS = value

Specifies the maximum number of simultaneous requests that are allowed to execute in the workload group. value must be 0 or a positive integer. The default setting for value, 0, allows unlimited requests. When the maximum concurrent requests are reached, a user in that group can sign in, but is placed in a wait state until concurrent requests are dropped below the value specified.

USING { pool_name | "default" }

Associates the workload group with the user-defined resource pool identified by pool_name, which in effect puts the workload group in the resource pool. If pool_name isn't provided or if the USING argument isn't used, the workload group is put in the predefined Resource Governor default pool.

The option "default" is case-sensitive and must be enclosed by quotation marks ("") or brackets ([]) when used with ALTER WORKLOAD GROUP to avoid conflict with DEFAULT, which is a system reserved word. For more information, see Database Identifiers.

Remarks

ALTER WORKLOAD GROUP is allowed on the default group.

Changes to the workload group configuration don't take effect until after ALTER RESOURCE GOVERNOR RECONFIGURE is executed. When changing a plan affecting setting, the new setting will only take effect in previously cached plans after executing DBCC FREEPROCCACHE (*pool_name*), where pool_name is the name of a Resource Governor resource pool on which the workload group is associated with.

  • If changing MAX_DOP to 1, executing DBCC FREEPROCCACHE isn't required because parallel plans can run in serial mode. However, it may not be as efficient as a plan compiled as a serial plan.

  • If changing MAX_DOP from 1 to 0 or a value greater than 1, executing DBCC FREEPROCCACHE isn't required. However, serial plans can't run in parallel, so clearing the respective cache will allow new plans to potentially be compiled using parallelism.

Caution

Clearing cached plans from a resource pool that is associated with more than one workload group will affect all workload groups with the user-defined resource pool identified by pool_name.

When executing DDL statements, you should be familiar with Resource Governor states. For more information, see Resource Governor.

REQUEST_MEMORY_GRANT_PERCENT: In SQL Server 2005 (9.x), index creation is allowed to use more workspace memory than initially granted for improved performance. This special handling is supported by Resource Governor in later versions, however, the initial grant and any additional memory grant are limited by resource pool and workload group settings.

Index creation on a partitioned table

The memory consumed by index creation on non-aligned partitioned table is proportional to the number of partitions involved. If the total required memory exceeds the per-query limit (REQUEST_MAX_MEMORY_GRANT_PERCENT) imposed by the Resource Governor workload group setting, this index creation may fail to execute. Because the "default" workload group allows a query to exceed the per-query limit with the minimum required memory to start for SQL Server 2005 (9.x) compatibility, the user may be able to run the same index creation in "default" workload group, if the "default" resource pool has enough total memory configured to run such query.

Permissions

Requires CONTROL SERVER permission.

Examples

The following example shows how to change the importance of requests in the default group from MEDIUM to LOW.

ALTER WORKLOAD GROUP "default"
WITH (IMPORTANCE = LOW);
GO
ALTER RESOURCE GOVERNOR RECONFIGURE;
GO

The following example shows how to move a workload group from the pool that it's in, to the default pool.

ALTER WORKLOAD GROUP adHoc
USING [default];
GO
ALTER RESOURCE GOVERNOR RECONFIGURE;
GO

See also

* Azure Synapse
Analytics *
 

 

Azure Synapse Analytics

Alters an existing workload group.

See the ALTER WORKLOAD GROUP behavior section below for further details on how ALTER WORKLOAD GROUP behaves on a system with running and queued requests.

Restrictions in place for CREATE WORKLOAD GROUP also apply to ALTER WORKLOAD GROUP. Prior to modifying parameters, query sys.workload_management_workload_groups to ensure the values are within acceptable ranges.

Syntax

ALTER WORKLOAD GROUP group_name
WITH
([ MIN_PERCENTAGE_RESOURCE = value ]
  [ [ , ] CAP_PERCENTAGE_RESOURCE = value ]
  [ [ , ] REQUEST_MIN_RESOURCE_GRANT_PERCENT = value ]
  [ [ , ] REQUEST_MAX_RESOURCE_GRANT_PERCENT = value ]
  [ [ , ] IMPORTANCE = { LOW | BELOW_NORMAL | NORMAL | ABOVE_NORMAL | HIGH }]
  [ [ , ] QUERY_EXECUTION_TIMEOUT_SEC = value ] )
  [ ; ]

Arguments

group_name

Is the name of the existing user-defined workload group being altered. group_name isn't alterable.

MIN_PERCENTAGE_RESOURCE = value

value is an integer range from 0 to 100. When altering MIN_PERCENTAGE_RESOURCE, the sum of MIN_PERCENTAGE_RESOURCE across all workload groups can't exceed 100. Altering MIN_PERCENTAGE_RESOURCE requires all running queries to complete in the workload group before the command will complete. For more information, see the ALTER WORKLOAD GROUP behavior section in this article.

CAP_PERCENTAGE_RESOURCE = value

value is an integer range from 1 through 100. The value for CAP_PERCENTAGE_RESOURCE must be greater than MIN_PERCENTAGE_RESOURCE. Altering CAP_PERCENTAGE_RESOURCE requires all running queries to complete in the workload group before the command will complete. For more information, see the ALTER WORKLOAD GROUP behavior section in this article.

REQUEST_MIN_RESOURCE_GRANT_PERCENT = value

value is a decimal with a range between 0.75 to 100.00. The value for REQUEST_MIN_RESOURCE_GRANT_PERCENT needs to be a factor of MIN_PERCENTAGE_RESOURCE and be less than CAP_PERCENTAGE_RESOURCE.

REQUEST_MAX_RESOURCE_GRANT_PERCENT = value

value is a decimal and must be greater than REQUEST_MIN_RESOURCE_GRANT_PERCENT.

IMPORTANCE = { LOW | BELOW_NORMAL | NORMAL | ABOVE_NORMAL | HIGH }

Alters the default importance of a request for the workload group.

QUERY_EXECUTION_TIMEOUT_SEC = value

Alters the maximum time, in seconds, that a query can execute before it's canceled. Value must be 0 or a positive integer. The default setting for value is 0, which means unlimited.

Permissions

Requires CONTROL DATABASE permission.

Example

The below example checks the values in the catalog view for a workload group named wgDataLoads, and changes the values.

SELECT *
FROM sys.workload_management_workload_groups
WHERE [name] = 'wgDataLoads'

ALTER WORKLOAD GROUP wgDataLoads WITH
( MIN_PERCENTAGE_RESOURCE            = 40
, CAP_PERCENTAGE_RESOURCE            = 80
, REQUEST_MIN_RESOURCE_GRANT_PERCENT = 10 )

ALTER WORKLOAD GROUP behavior

At any point in time there are three types of requests in the system:

  • Requests that haven't been classified yet.
  • Requests that are classified, and waiting, for object locks or system resources.
  • Requests that are classified, and running.

Based on the properties of a workload group being altered, the timing of when the settings take effect will differ.

Importance or query_execution_timeout

For the importance and query_execution_timeout properties, non-classified requests pick up the new config values. Waiting and running requests execute with the old configuration. The ALTER WORKLOAD GROUP request executes immediately regardless if there are running queries in the workload group.

REQUEST_MIN_RESOURCE_GRANT_PERCENT or REQUEST_MAX_RESOURCE_GRANT_PERCENT

For REQUEST_MIN_RESOURCE_GRANT_PERCENT and REQUEST_MAX_RESOURCE_GRANT_PERCENT, running requests execute with the old configuration. Waiting requests and non-classified requests pick up the new config values. The ALTER WORKLOAD GROUP request executes immediately regardless if there are running queries in the workload group.

MIN_PERCENTAGE_RESOURCE or CAP_PERCENTAGE_RESOURCE

For MIN_PERCENTAGE_RESOURCE and CAP_PERCENTAGE_RESOURCE, running requests execute with the old configuration. Waiting requests and non-classified requests pick up the new config values.

Changing MIN_PERCENTAGE_RESOURCE and CAP_PERCENTAGE_RESOURCE requires draining of running requests in the workload group that is being altered. When decreasing MIN_PERCENTAGE_RESOURCE, the freed resources are returned to the share pool allowing requests from other workload groups the ability to utilize. Conversely, increasing the MIN_PERCENTAGE_RESOURCE will wait until requests utilizing only the needed resources from the shared pool to complete. The ALTER WORKLOAD GROUP operation will have prioritized access to shared resources over other requests waiting to be executed on shared pool. If the sum of MIN_PERCENTAGE_RESOURCE exceeds 100%, the ALTER WORKLOAD GROUP request fails immediately.

Locking behavior

Altering a workload group requires a global lock across all workload groups. A request to alter a workload group would queue behind already submitted create or drop workload group requests. If a batch of alter statements is submitted at once, they're processed in the order in which they're submitted.

See also