Updated on 2024-08-20 GMT+08:00

SQL Statement Concurrency Control Function

  • gs_add_workload_rule(rule_type, rule_name, databases, start_time, end_time, max_workload, option_val)

    Description: Creates an SQL statement concurrency control rule. Users must have the sysadmin permission.

    Parameters: For details, see Table 1.

    Return type: int8

    Table 1 gs_add_workload_rule parameters

    Parameter

    Type

    Description

    Value Range

    rule_type

    text

    Type of the concurrency control rule, which is case insensitive.

    "sqlid": Concurrency control is based on the unique SQL ID.

    "select", "insert", "update", "delete", and "merge": Concurrency control is based on the query type and keyword.

    "resource": Instance-level concurrency control is based on the system resource usage.

    rule_name

    name

    Name of a concurrency control rule, which is used to search for the concurrency control rule.

    Any character string or NULL.

    databases

    name[]

    Array of database names for which the concurrency control rule takes effect. The value is case-sensitive.

    List of names of created databases. The value can be NULL, indicating that the configuration takes effect in all databases.

    Currently, the database list takes effect only when rule_type is set to a query type because a unique SQL ID is bound to a database and belongs to only one database. The concurrency control rules based on resource usage take effect for instances, that is, all databases.

    start_time

    timestamptz

    Start time of a concurrency control rule.

    The value can be NULL, indicating that it takes effect from now on.

    end_time

    timestamptz

    End time of a concurrency control rule.

    The value can be NULL, indicating that the rule is always effective.

    max_workload

    int8

    Maximum number of concurrent requests set in a concurrency control rule.

    -

    option_val

    text[]

    Supplementary information about the concurrency control rule.

    It matches rule_type. The matching relationship is as follows:

    • "sqlid": specifies the unique ID of the SQL statement whose concurrency is to be controlled and slow SQL control rule. The format is '{id=1234, time_limit=100, max_execute_time=500, max_iops=1}', in which id indicates the unique SQL ID and is mandatory. You can obtain it from the dbe_perf.statement or pg_stat_activity view. Others are optional. For details about their meanings, see section "Hint for Setting Slow SQL Control Rules."
    • "select", "insert", "update", "delete", and "merge": keyword sequence for concurrency control, which is case insensitive and can be NULL.
    • "resource": resource threshold for triggering instance-level concurrency control. The value is in the format of '{cpu-80, memory-70}', indicating the OS resource threshold for triggering instance-level concurrency control. The value can be NULL, indicating that concurrency control is performed regardless of the resource usage.

    Example:

     1
     2
     3
     4
     5
     6
     7
     8
     9
    10
    11
    12
    13
    14
    15
    16
    17
    gaussdb=#  SELECT gs_add_workload_rule('sqlid', 'rule for one query', '', now(), '', 20, '{id=32413214}');
     gs_add_workload_rule 
    ----------------------
                       1
    (1 row)
    gaussdb=#  CREATE database db1;
    gaussdb=#  CREATE database db2;
    gaussdb=#  SELECT gs_add_workload_rule('select', 'rule for select', '{db1, db2}', '', '', 100, '{tb1, tb2}');
     gs_add_workload_rule 
    ----------------------
                       2
    (1 row)
    gaussdb=#  SELECT gs_add_workload_rule('resource', 'rule for resource', '{}', '', '', 20, '{cpu-80}');
     gs_add_workload_rule 
    ----------------------
                       3
    (1 row)
    
  • gs_update_workload_rule(rule_id, rule_name, databases, start_time, end_time, max_workload, option_val)

    Description: To update an SQL statement concurrency control rule, users need to reset all parameters instead of only some parameters. Users must have the sysadmin permission.

    Parameters: For details, see Table 2.

    Return type: Boolean

    Table 2 gs_update_workload_rule parameters

    Parameter

    Type

    Description

    Value Range

    rule_id

    int8

    ID of the concurrency control rule to be updated.

    -

    rule_name

    name

    Name of a concurrency control rule, which is used to search for the concurrency control rule.

    Any character string or NULL.

    databases

    name[]

    Array of database names for which the concurrency control rule takes effect. The value is case sensitive.

    List of names of created databases. The value can be NULL, indicating that the configuration takes effect in all databases.

    Currently, the database list takes effect only when rule_type is set to a query type because a unique SQL ID is bound to a database and belongs to only one database. The concurrency control rules based on resource usage take effect for instances, that is, all databases.

    start_time

    timestamptz

    Start time of a concurrency control rule.

    The value can be NULL, indicating that it takes effect from now on.

    end_time

    timestamptz

    End time of a concurrency control rule.

    The value can be NULL, indicating that the rule is always effective.

    max_workload

    int8

    Maximum number of concurrent requests set in a concurrency control rule.

    -

    option_val

    text[]

    Supplementary information about the concurrency control rule.

    It matches rule_type. The matching relationship is as follows:

    • "sqlid": specifies the unique ID of the SQL statement whose concurrency is to be controlled and slow SQL control rule. The format is '{id=1234, time_limit=100, max_execute_time=500, max_iops=1}', in which id indicates the unique SQL ID and is mandatory. You can obtain it from the dbe_perf.statement or pg_stat_activity view. Others are optional. For details about their meanings, see section "Hint for Setting Slow SQL Control Rules."
    • "select", "insert", "update", "delete", and "merge": keyword sequence for concurrency control, which is case insensitive and can be NULL.
    • "resource": resource threshold for triggering instance-level concurrency control. The value is in the format of '{cpu-80, memory-70}', indicating the OS resource threshold for triggering instance-level concurrency control. The value can be NULL, indicating that concurrency control is performed regardless of the resource usage.

    Example:

    1
    2
    3
    4
    5
    6
    gaussdb=#  CREATE database db1;
    gaussdb=#  SELECT gs_update_workload_rule(2, 'rule for select 2', '{db1}', now(), '', 50, '{tb1}');
     gs_update_workload_rule 
    -------------------------
     t
    (1 row)
    
  • gs_delete_workload_rule(rule_id int8)

    Description: Deletes an SQL statement concurrency control rule. Users must have the sysadmin permission.

    Parameter: rule_id indicates the ID of the concurrency control rule to be updated. The type is int8.

    Return type: Boolean

    Example:

    1
    2
    3
    4
    5
    gaussdb=#  SELECT gs_delete_workload_rule(3);
     gs_delete_workload_rule 
    -------------------------
     t
    (1 row)
    
  • gs_get_workload_rule_stat(rule_id)

    Description: Queries the number of times that SQL statements are blocked by SQL statement concurrency control rules. Users must have the sysadmin permission.

    Parameter: rule_id indicates the ID of the concurrency control rule to be queried. The type is int8. You can set rule_id to –1, indicating that all SQL statement concurrency control rules are queried.

    Table 3 Return value types

    Name

    Type

    Description

    rule_id

    int8

    ID of the SQL statement concurrency control rule.

    validate_count

    int8

    Number of SQL statements intercepted by the SQL statement concurrency control rule.

    Example:

     1
     2
     3
     4
     5
     6
     7
     8
     9
    10
    11
    gaussdb=#  SELECT * FROM gs_get_workload_rule_stat(1);
     rule_id | validate_count 
    ---------+----------------
           1 |              0
    (1 row)
    gaussdb=#  SELECT * FROM gs_get_workload_rule_stat(-1);
     rule_id | validate_count 
    ---------+----------------
           1 |              0
           2 |              0
    (2 rows)