Creates a user-defined resource pool.

Syntax

CREATE RESOURCE POOL pool-name [ FOR subcluster ] [ parameter-name setting ]...

Arguments

pool-name

Name of the resource pool. If you specify a resource pool name with uppercase letters, Vertica converts them to lowercase letters.

The following naming requirements apply:

• Built-in pool names cannot be used for user-defined pools.

• All user-defined global resource pools must have unique names.

• Names of global and subcluster-assigned resource pools cannot be the same. However, the same name can be shared by resource pools of different subclusters.

FOR subcluster

Eon Mode only, the subcluster to associate with this resource pool, where subcluster is one of the following:

• SUBCLUSTER subcluster-name: Resource pool for an existing subcluster. You cannot be connected to this subcluster, otherwise Vertica returns an error.
• CURRENT SUBCLUSTER: Resource pool for the subcluster that you are connected to.

If omitted, the resource pool is created globally.

parameter-name setting

A resource pool parameter and its initial value. If you omit this argument, Vertica sets this resource pool's parameters to their default values (see Parameters).

Parameters

Default values specified here pertain only to user-defined resource pools. For built-in pool default values, see Built-in resource pools configuration, or query system table RESOURCE_POOL_DEFAULTS.

CASCADE TO

Secondary resource pool for executing queries that exceed the RUNTIMECAP setting of their assigned resource pool:

CASCADE TO secondary-pool

CPUAFFINITYMODE

Specifies whether the resource pool has exclusive or shared use of the CPUs specified in CPUAFFINITYSET:

CPUAFFINITYMODE { SHARED | EXCLUSIVE | ANY }

• SHARED: Queries that run in this resource pool share its CPUAFFINITYSET CPUs with other Vertica resource pools.
• EXCLUSIVE: Dedicates CPUAFFINITYSET CPUs to this resource pool only, and excludes other Vertica resource pools. If CPUAFFINITYSET is set as a percentage, then that percentage of CPU resources available to Vertica is assigned solely for this resource pool.
• ANY: Queries in this resource pool can run on any CPU, invalid if CPUAFFINITYSET designates CPU resources.

Default: ANY

CPUAFFINITYSET

CPUs available to this resource pool. All cluster nodes must have the same number of CPUs. The CPU resources assigned to this set are unavailable to general resource pools.

CPUAFFINITYSET {
  'cpu-index[,...]'
| 'cpu-indexi-cpu-indexn'
| 'integer%'
| NONE
}

• cpu-index[,...]: Dedicates one or more comma-delimited CPUs to this resource pool.
• cpu-indexi-cpu-indexn: Dedicates a range of contiguous CPU indexes i through n to this resource pool.
• integer%: Percentage of all available CPUs to use for this resource pool. Vertica rounds this percentage down to include whole CPU units.
• NONE (empty string): No affinity set is assigned to this resource pool. Queries associated with this pool are executed on any CPU.

Default: NONE

Important

CPUAFFINITYSET and CPUAFFINITYMODE must be set together in the same statement.

EXECUTIONPARALLELISM

Number of threads used to process any single query issued in this resource pool.

EXECUTIONPARALLELISM { limit | AUTO }

• limit: An integer value between 1 and the number of cores. Setting this parameter to a reduced value increases throughput of short queries issued in the resource pool, especially if queries are executed concurrently.
• AUTO or 0: Vertica calculates the setting from the number of cores, available memory, and amount of data in the system. Unless memory is limited, or the amount of data is very small, Vertica sets this parameter to the number of cores on the node.

Default: AUTO

MAXCONCURRENCY

Maximum number of concurrent execution slots available to the resource pool across the cluster:

MAXCONCURRENCY { integer | NONE }

NONE (empty string): Unlimited number of concurrent execution slots.

Default: NONE

MAXMEMORYSIZE

Maximum size per node the resource pool can grow by borrowing memory from the GENERAL pool:

MAXMEMORYSIZE {
  'integer%'
  |'integer{K|M|G|T}'
  NONE
}

• integer%: Percentage of total memory
• integer{K|M|G|T}: Amount of memory in kilobytes, megabytes, gigabytes, or terabytes
• NONE (empty string): Unlimited, resource pool can borrow any amount of available memory from the GENERAL pool.

Default: NONE

MAXQUERYMEMORYSIZE

Maximum amount of memory this resource pool can allocate at runtime to process a query. If the query requires more memory than this setting, Vertica stops execution and returns an error.

Set this parameter as follows:

MAXQUERYMEMORYSIZE {
  'integer%'
| 'integer{K|M|G|T}'
| NONE
}

• integer%: Percentage of MAXMEMORYSIZE for this resource pool.
• integer{K|M|G|T}: Amount of memory in kilobytes, megabytes, gigabytes, or terabytes, up to the value of MAXMEMORYSIZE.
• NONE (empty string): Unlimited; resource pool can borrow any amount of available memory from the GENERAL pool, within the limits set by MAXMEMORYSIZE.

Default: NONE

MEMORYSIZE

Total per-node memory available to the Vertica resource manager that is allocated to this resource pool:

MEMORYSIZE {
  'integer%'
| 'integer{K|M|G|T}'
}

• integer%: Percentage of total memory
• integer{K|M|G|T}: Amount of memory in kilobytes, megabytes, gigabytes, or terabytes

Default: 0%. No memory allocated, the resource pool borrows memory from the GENERAL pool.

PLANNEDCONCURRENCY

Preferred number of queries to execute concurrently in the resource pool. This setting applies to the entire cluster:

PLANNEDCONCURRENCY { num-queries | AUTO }

• num-queries: Integer value ≥ 1, the preferred number of queries to execute concurrently in the resource pool. When possible, query resource budgets are limited to allow this level of concurrent execution.

• AUTO: Value is calculated automatically at query runtime. Vertica sets this parameter to the lower of these two calculations, but never less than 4:

  • Number of logical cores

  • Memory divided by 2GB

  If the number of logical cores on each node is different, AUTO is calculated differently for each node. Distributed queries run like the minimal effective planned concurrency. Single node queries run with the planned concurrency of the initiator.

Default: AUTO

Tip

Change this parameter only after evaluating performance over a period of time.

PRIORITY

Priority of queries in this resource pool when they compete for resources in the GENERAL pool:

PRIORITY { integer | HOLD }

• integer: Negative or positive integer value, where higher numbers denote higher priority:

  • User-defined resource pool: -100 to 100

  • Built-in resource pools SYSQUERY, RECOVERY, and TM: -110 to 110

• HOLD: Sets priority to -999. Queries in this resource pool are queued until QUEUETIMEOUT is reached.

Default: 0

QUEUETIMEOUT

Maximum time a request can wait for pool resources before it is rejected, not more than one year:

QUEUETIMEOUT { integer | 'interval' | 'NONE' }

• integer: Maximum wait time in seconds

• [interval](/en/sql-reference/language-elements/literals/datetime-literals/interval-literal/): Maximum wait time expressed in the following format:

  num year num months num [days] HH:MM:SS.ms
  

• NONE (empty string): No maximum wait time, request can be queued indefinitely, up to one year.

If the value that you specify resolves to more than one year, Vertica returns with a warning and sets the parameter to 365 days:

=> ALTER RESOURCE POOL user_0 QUEUETIMEOUT '11 months 50 days 08:32';
WARNING 5693:  Using 1 year for QUEUETIMEOUT
ALTER RESOURCE POOL
=> SELECT QUEUETIMEOUT FROM resource_pools WHERE name = 'user_0';
 QUEUETIMEOUT
--------------
 365
(1 row)

Default: 00:05 (5 minutes)

RUNTIMECAP

Maximum execution time allowed to queries in this resource pool, not more than one year, otherwise Vertica returns with an error. If a query exceeds this setting, it tries to cascade to a secondary pool:

RUNTIMECAP { 'interval' | NONE }

• interval: Maximum wait time expressed in the following format:

  num year num month num [day] HH:MM:SS.ms
  

• NONE (empty string): No maximum wait time, request can be queued indefinitely, up to one year.

If the user or session also has a RUNTIMECAP, the shorter limit applies.

RUNTIMEPRIORITY

Determines how the resource manager should prioritize dedication of run-time resources (CPU, I/O bandwidth) to queries already running in this resource pool:

RUNTIMEPRIORITY { HIGH | MEDIUM | LOW }

Default: MEDIUM

RUNTIMEPRIORITYTHRESHOLD

Maximum time (in seconds) in which query processing must complete before the resource manager assigns to it the resource pool's RUNTIMEPRIORITY. All queries begin execution with a priority of HIGH.

RUNTIMEPRIORITYTHRESHOLD seconds

Default: 2

SINGLEINITIATOR

Set to false for backward compatibility. Do not change this setting.

Privileges

Superuser

Examples

This example shows how to create a resource pool with MEMORYSIZE of 1800 MB.

=> CREATE RESOURCE POOL ceo_pool MEMORYSIZE '1800M' PRIORITY 10;
CREATE RESOURCE POOL

Assuming the CEO report user already exists, associate this user with the preceding resource pool using ALTER USER statement.

=> GRANT USAGE ON RESOURCE POOL ceo_pool to ceo_user;
GRANT PRIVILEGE
=> ALTER USER ceo_user RESOURCE POOL ceo_pool;
ALTER USER

Issue the following command to confirm that the ceo_user is associated with the ceo_pool:

=> SELECT * FROM users WHERE user_name ='ceo_user';
-[ RECORD 1 ]-----+--------------------------------------------------
user_id           | 45035996273733402
user_name         | ceo_user
is_super_user     | f
profile_name      | default
is_locked         | f
lock_time         |
resource_pool     | ceo_pool
memory_cap_kb     | unlimited
temp_space_cap_kb | unlimited
run_time_cap      | unlimited
all_roles         |
default_roles     |
search_path       | "$user", public, v_catalog, v_monitor, v_internal

This exampleshows how to create and designate secondary resource pools.

=> CREATE RESOURCE POOL rp3 RUNTIMECAP '5 minutes';
=> CREATE RESOURCE POOL rp2 RUNTIMECAP '3 minutes' CASCADE TO rp3;
=> CREATE RESOURCE POOL rp1 RUNTIMECAP '1 minute' CASCADE TO rp2;
=> SET SESSION RESOURCE_POOL = rp1;

This Eon Mode example confirms the current subcluster name, then creates a resource pool for the current subcluster:

=> SELECT CURRENT_SUBCLUSTER_NAME();
 CURRENT_SUBCLUSTER_NAME
-------------------------
 analytics_1
(1 row)

=> CREATE RESOURCE POOL dashboard FOR SUBCLUSTER analytics_1;
CREATE RESOURCE POOL

See also

• ALTER RESOURCE POOL
• CREATE USER
• DROP RESOURCE POOL
• SET SESSION RESOURCE_POOL
• SET SESSION MEMORYCAP
• Managing workloads