diff --git a/gpdb-doc/dita/admin_guide/workload_mgmt_resgroups.xml b/gpdb-doc/dita/admin_guide/workload_mgmt_resgroups.xml index 0ba491273ebcd5c1f6b69bb4dd01f9ce3806819d..769a007bdeed8a38cda9dabddda6b4580cc250e5 100644 --- a/gpdb-doc/dita/admin_guide/workload_mgmt_resgroups.xml +++ b/gpdb-doc/dita/admin_guide/workload_mgmt_resgroups.xml @@ -50,13 +50,14 @@
    • +
    • +
    • -
    • @@ -228,6 +229,14 @@ +

    + For queries managed by resource groups that are configured to use the + vmtracker memory auditor, Greenplum Database supports the automatic + termination of queries based on the amount of memory the queries are using. See the server + configuration parameter . +

    @@ -547,7 +556,7 @@ SET statement_mem='10 MB';

     - Using Resource Groups + Configuring and Using Resource Groups Significant Greenplum Database performance degradation has been observed when enabling resource group-based workload management on RedHat 6.x and CentOS 6.x @@ -822,6 +831,24 @@ gpstart

    + + Configuring Automatic Query Termination + +

    When resource groups have a global shared memory pool, the server configuration parameter + sets the percent of utilized global shared memory that + triggers the termination of queries that are managed by resource groups that are configured + to use the vmtracker memory auditor, such as admin_group + and default_group.

    +

    Resource groups have a global shared memory pool when the sum of the + MEMORY_LIMIT attribute values configured for all resource groups is less + than 100. For example, if you have 3 resource groups configured with + MEMORY_LIMIT values of 10 , 20, and 30, then global shared memory is 40% + = 100% - (10% + 20% + 30%).

    +

    For information about global shared memory, see .

    + +     Assigning a Resource Group to a Role diff --git a/gpdb-doc/dita/ref_guide/config_params/guc-list.xml b/gpdb-doc/dita/ref_guide/config_params/guc-list.xml index 74d5157f3d9b4ab6b52cb032880c07b14c26571e..cb1534fd84cd637bab196b799c473db596495f0f 100644 --- a/gpdb-doc/dita/ref_guide/config_params/guc-list.xml +++ b/gpdb-doc/dita/ref_guide/config_params/guc-list.xml @@ -4328,7 +4328,8 @@ gp_resource_manager

    Identifies the resource management scheme currently enabled in the Greenplum Database - cluster. The default scheme is to use resource queues.

    + cluster. The default scheme is to use resource queues. For information about Greenplum + Database resource management, see .

    @@ -8001,20 +8002,49 @@ runaway_detector_activation_percent - The runaway_detector_activation_percent server configuration parameter - is enforced only when resource queue-based resource management is active. -

    Sets the percentage of Greenplum Database vmem memory that triggers the termination of - queries. If the percentage of vmem memory that is utilized for a Greenplum Database segment - exceeds the specified value, Greenplum Database terminates queries based on memory usage, - starting with the query consuming the largest amount of memory. Queries are terminated until - the percentage of utilized vmem is below the specified percentage.

    +

    For queries that are managed by resource queues or resource groups, this parameter + determines when Greenplum Database terminates running queries based on the amount of memory + the queries are using. A value of 100 disables the automatic termination of queries based on + the percentage of memory that is utilized.

    +

    Either the resource queue or the resource group management scheme can be active in + Greenplum Database; both schemes cannot be active at the same time. The server configuration + parameter controls which scheme is + active.

    +

    When resource queues are enabled - This parameter sets the percent of utilized + Greenplum Database vmem memory that triggers the termination of queries. If the percentage + of vmem memory that is utilized for a Greenplum Database segment exceeds the specified + value, Greenplum Database terminates queries managed by resource queues based on memory + usage, starting with the query consuming the largest amount of memory. Queries are + terminated until the percentage of utilized vmem is below the specified percentage.

    Specify the maximum vmem value for active Greenplum Database segment instances with the - server configuration parameter .

    -

    For example, if vmem memory is set to 10GB, and the value of - runaway_detector_activation_percent is 90 (90%), Greenplum Database - starts terminating queries when the utilized vmem memory exceeds 9 GB.

    -

    A value of 0 disables the automatic termination of queries based on percentage of vmem that - is utilized.

    + server configuration parameter .

    +

    For example, if vmem memory is set to 10GB, and this parameter is 90 (90%), Greenplum + Database starts terminating queries when the utilized vmem memory exceeds 9 GB.

    +

    For information about resource queues, see .

    +

    When resource groups are enabled - This parameter sets the percent of utilized + resource group global shared memory that triggers the termination of queries that are + managed by resource groups that are configured to use the vmtracker memory + auditor, such as admin_group and default_group. For + information about memory auditors, see .

    +

    Resource groups have a global shared memory pool when the sum of the + MEMORY_LIMIT attribute values configured for all resource groups is less + than 100. For example, if you have 3 resource groups configured with + memory_limit values of 10 , 20, and 30, then global shared memory is 40% + = 100% - (10% + 20% + 30%). See .

    +

    If the percentage of utilized global shared memory exceeds the specified value, Greenplum + Database terminates queries based on memory usage, selecting from queries managed by the + resource groups that are configured to use the vmtracker memory auditor. + Greenplum Database starts with the query consuming the largest amount of memory. Queries are + terminated until the percentage of utilized global shared memory is below the specified + percentage.

    +

    For example, if global shared memory is 10GB, and this parameter is 90 (90%), Greenplum + Database starts terminating queries when the utilized global shared memory exceeds 9 GB.

    +

    For information about resource groups, see .

    diff --git a/gpdb-doc/dita/ref_guide/config_params/guc_category-list.xml b/gpdb-doc/dita/ref_guide/config_params/guc_category-list.xml index c71a909d9200a52593c0c13ff0ed416f9b35b82c..ab7c52d1beaa3407bfd132d5c3a535331685bc6e 100644 --- a/gpdb-doc/dita/ref_guide/config_params/guc_category-list.xml +++ b/gpdb-doc/dita/ref_guide/config_params/guc_category-list.xml @@ -1190,6 +1190,10 @@

    memory_spill_ratio

    +

    + runaway_detector_activation_percent +

    statement_mem

    diff --git a/gpdb-doc/dita/ref_guide/sql_commands/CREATE_RESOURCE_GROUP.xml b/gpdb-doc/dita/ref_guide/sql_commands/CREATE_RESOURCE_GROUP.xml index 08cc4068b97f6c86fc1a45255ea4c96c137f1a85..6f256c80565eac4f30e65dfd77c17f99dc551d12 100644 --- a/gpdb-doc/dita/ref_guide/sql_commands/CREATE_RESOURCE_GROUP.xml +++ b/gpdb-doc/dita/ref_guide/sql_commands/CREATE_RESOURCE_GROUP.xml @@ -88,8 +88,17 @@ MEMORY_AUDITOR {vmtracker | cgroup} - The memory auditor for the resource group. Greenplum Database employs virtual memory tracking for role resources and cgroup memory tracking for resources used by external components. The default MEMORY_AUDITOR is vmtracker. When you create a resource group with vmtracker memory auditing, Greenplum Database tracks that resource group's memory internally. - When you create a resource group specifying the cgroup MEMORY_AUDITOR, Greenplum Database defers the accounting of memory used by that resource group to cgroups. CONCURRENCY must be zero (0) for a resource group that you create for external components such as PL/Container. You cannot assign a resource group that you create for external components to a Greenplum Database role. + The memory auditor for the resource group. Greenplum Database employs virtual memory + tracking for role resources and cgroup memory tracking for resources used by external + components. The default MEMORY_AUDITOR is vmtracker. + When you create a resource group with vmtracker memory auditing, + Greenplum Database tracks that resource group's memory internally. + When you create a resource group specifying the cgroup + MEMORY_AUDITOR, Greenplum Database defers the accounting of memory used + by that resource group to cgroups. CONCURRENCY must be zero (0) for a + resource group that you create for external components such as PL/Container. You cannot + assign a resource group that you create for external components to a Greenplum Database + role.