Snowpark Container Services: Working with compute pools¶

Creating a compute pool¶

A compute pool is an account-level construct, analogous to a Snowflake virtual warehouse. The naming scope of the compute pool is your account. That is, you cannot have multiple compute pools with the same name in your account.

The minimum information required to create a compute pool includes the following:

• The machine type (referred to as the instance family) to provision for the compute pool nodes

• The minimum nodes to launch the compute pool with

• The maximum number of nodes the compute pool can scale to (Snowflake manages the scaling.)

If you expect a substantial load or sudden bursts of activity on the services you intend to run within your compute pool, you can set a minimum node count greater than 1. This approach ensures that additional nodes are readily available when needed, instead of waiting for autoscaling to start.

Setting a maximum node limit prevents an unexpectedly large number of nodes from being added to your compute pool by Snowflake autoscaling. This can be crucial in scenarios such as unexpected load spikes or issues in your code that might cause Snowflake to allocate a larger number of compute pool nodes than originally planned.

To create a compute pool using Snowsight, or SQL:

Snowsight:

1. Select Admin » Compute pools.

2. Select your username at the bottom of the navigation bar and switch to the ACCOUNTADMIN role, or any role that is allowed to create a compute pool.

3. Select + Compute Pool.

4. In the New compute pool UI, specify the required information (the compute pool name, the instance family, and the node limit).

5. Select Create Compute Pool.

SQL:

Execute the CREATE COMPUTE POOL command.

For example, the following command creates a one-node compute pool:

CREATE COMPUTE POOL tutorial_compute_pool
  MIN_NODES = 1
  MAX_NODES = 1
  INSTANCE_FAMILY = CPU_X64_XS;

Copy

The instance family identifies the type of machine you want to provision for computer nodes in the compute pool. Specifying instance family in creating a compute pool is similar to specifying warehouse size (XSMALL, SMALL, MEDIUM, LARGE and so on) when creating a warehouse. The following table lists the available machine types.

INSTANCE_FAMILY, Snowflake Service Consumption Table Mapping	vCPU	Memory (GiB)	Storage (GiB)	Bandwidth limit (Gbps)	GPU	GPU Memory per GPU (GiB)	Node limit	Description
CPU_X64_XS, . CPU | XS	1	6	100	Up to 12.5	n/a	n/a	50	Smallest instance available for Snowpark Containers. Ideal for cost-savings and getting started.
CPU_X64_S, . CPU | S	3	13	100	Up to 12.5	n/a	n/a	50	Ideal for hosting multiple services/jobs while saving cost.
CPU_X64_M, . CPU | M	6	28	100	Up to 12.5	n/a	n/a	50	Ideal for having a full stack application or multiple services
CPU_X64_L, . CPU | L	28	116	100	12.5	n/a	n/a	50	For applications which need an unusually large number of CPUs, memory and Storage.
HIGHMEM_X64_S, . High-Memory CPU | S	6	58	100	AWS: Up to 12.5, Azure: 8	n/a	n/a	50	For memory intensive applications.
HIGHMEM_X64_M, . High-Memory CPU | M . (AWS only)	28	240	100	12.5	n/a	n/a	50	For hosting multiple memory intensive applications on a single machine.
HIGHMEM_X64_M, . High-Memory CPU | M . (Azure only)	28	244	100	16	n/a	n/a	50	For hosting multiple memory intensive applications on a single machine.
HIGHMEM_X64_L, . High-Memory CPU | L . (AWS only)	124	984	100	50	n/a	n/a	20	Largest AWS high-memory machine available for processing large in-memory data.
HIGHMEM_X64_SL, . High-Memory CPU | L . (Azure only)	92	654	100	32	n/a	n/a	20	Largest Azure high-memory machine available for processing large in-memory data.
GPU_NV_S, . GPU | S . (AWS only, except Paris and Osaka regions)	6	27	100	Up to 10	1 NVIDIA A10G	24	10	Our smallest NVIDIA GPU size available for Snowpark Containers to get started.
GPU_NV_M, . GPU | M . (AWS only, except not Paris and Osaka regions)	44	178	100	40	4 NVIDIA A10G	24	10	Optimized for intensive GPU usage scenarios like Computer Vision or LLMs/VLMs.
GPU_NV_L, . GPU | L . (AWS only, available only in AWS US West and US East regions by request; limited availability might be possible in other regions upon request)	92	1112	100	400	8 NVIDIA A100	40	On request	Largest GPU instance for specialized and advanced GPU cases like LLMs and Clustering, etc.
GPU_NV_XS, . GPU | XS . (Azure only, except not Switzerland North and UAE North regions)	3	26	100	8	1 NVIDIA T4	16	10	Our smallest Azure NVIDIA GPU size available for Snowpark Containers to get started.
GPU_NV_SM, . GPU | SM . (Azure only, except not Central US region)	32	424	100	40	1 NVIDIA A10	24	10	A smaller Azure NVIDIA GPU size available for Snowpark Containers to get started.
GPU_NV_2M, . GPU | 2M . (Azure only, except not Central US region)	68	858	100	80	2 NVIDIA A10	24	5	Optimized for intensive GPU usage scenarios like Computer Vision or LLMs/VLMs.
GPU_NV_3M, . GPU | 3M . (Azure only, except not North Europe and UAE North regions)	44	424	100	40	2 NVIDIA A100	80	On request	Optimized for memory-intensive GPU usage scenarios like Computer Vision or LLMs/VLMs.
GPU_NV_SL, . GPU | SL . (Azure only, except not North Europe and UAE North regions)	92	858	100	80	4 NVIDIA A100	80	On request	Largest GPU instance for specialized and advanced GPU cases like LLMs and Clustering, etc.

For information about available instance families, see CREATE COMPUTE POOL.

Managing a compute pool¶

You can manage a compute pool using Snowsight, or SQL.

In Snowsight, you choose the more option (…) next to the compute pool name, and choose the desired operation from the menu. The section explains SQL commands you can use to manage a compute pool.

Snowpark Container Services provides the following commands to manage compute pools:

• Monitoring: Use the SHOW COMPUTE POOLS command to get information about compute pools.

• Operating: Use the ALTER COMPUTE POOL command to change the state of a compute pool.

  ALTER COMPUTE POOL <name> { SUSPEND | RESUME | STOP ALL }
  

  Copy

  When you suspend a compute pool, Snowflake suspends all services except the job services. The job services continue to run until they reach a terminal state (DONE or FAILED), after which the compute pool nodes are released.

  A suspended compute pool must be resumed before you can start a new service. If the compute pool is configured to auto-resume (with the AUTO_RESUME property set to TRUE), Snowflake automatically resumes the pool when a service is submitted to it. Otherwise, you need to run the ALTER COMPUTE POOL command to manually resume the compute pool.

• Modifying: Use the ALTER COMPUTE POOL command to change compute pool properties.

  ALTER COMPUTE POOL <name> SET propertiesToAlter = <value>
  propertiesToAlter := { MIN_NODES | MAX_NODES | AUTO_RESUME | AUTO_SUSPEND_SECS | COMMENT }
  

  Copy

  When you decrease MAX_NODES, note the following potential effects:

  • Snowflake might need to terminate one or more service instances and restart them on other available nodes in the compute pool. If MAX_NODES is set too low, Snowflake might be unable to schedule certain service instances.

  • If the node terminated had a job service execution in progress, the job execution will fail. Snowflake will not restart the job service.

    Example:

    ALTER COMPUTE POOL MYPOOL SET MIN_NODES = 2  MAX_NODES = 2;
    

    Copy

• Removing: Use the DROP COMPUTE POOL command to remove a compute pool.

  Example:

  DROP COMPUTE POOL <name>
  

  Copy

  You must stop all running services before you can drop a compute pool.

• Listing compute pools and viewing properties: Use SHOW COMPUTE POOLS and DESCRIBE COMPUTE POOL commands. For examples, see Show Compute Pools.

Compute pool lifecycle¶

A compute pool can be in any of the following states:

• IDLE: The compute pool has the desired number of virtual machine (VM) nodes, but no services are scheduled. In this state, autoscaling can shrink the compute pool to the minimum size due to lack of activity.

• ACTIVE: The compute pool has at least one service running or scheduled to run on it. The pool can grow (up to the maximum nodes) or shrink (down to the minimum nodes) in response to load or user actions.

• SUSPENDED: The pool currently contains no running virtual machine nodes, but if the AUTO_RESUME compute pool property is set to TRUE, the pool will automatically resume when a service is scheduled.

The following states are transient:

• STARTING: When you create or resume a compute pool, the compute pool enters the STARTING state until at least one node is provisioned.

• STOPPING: When you suspend a compute pool (using ALTER COMPUTE POOL), the compute pool enters the STOPPING state until Snowflake has released all nodes in the compute pool. When you suspend a compute pool, Snowflake suspends all services except the job services. The job services continue to run until they reach a terminal state (DONE or FAILED), after which the compute pool nodes are released.

• RESIZING: When you create a compute pool, initially it enters the STARTING state. After it has one node provisioned, it enters the RESIZING state until the minimum number of nodes (as specified in CREATE COMPUTE POOL) are provisioned. When you change a compute pool (ALTER COMPUTE POOL) and update the minimum and maximum node values, the pool enters the RESIZING state until the minimum nodes are provisioned. Note that autoscaling of a compute pool also puts the compute pool in the RESIZING state.

Compute pool privileges¶

When you work with compute pools, the following privilege model applies:

• To create a compute pool in an account, the current role needs the CREATE COMPUTE POOL privilege on the account. If you create a pool, as an owner you have OWNERSHIP permission, which grants full control over that compute pool. Having OWNERSHIP of one compute pool doesn’t imply any permissions on other compute pools.

• For compute pool management, the following privileges (capabilities) are supported:

  Privilege	Usage
  MODIFY	Enables altering any compute pool properties, including changing the size.
  MONITOR	Enables viewing compute pool usage, including describing compute pool properties.
  OPERATE	Enables changing the state of the compute pool (suspend, resume). In addition, enables stopping any scheduled services (including job services).
  USAGE	Enables creating services in the compute pool. Note that when a compute pool is in a suspended state and has its AUTO_RESUME property set to true, a role with USAGE permission on the compute pool can implicitly trigger the compute pool’s resumption when they start or resume a service, even if the role lacks the OPERATE permission.
  OWNERSHIP	Grants full control over the compute pool. Only a single role can hold this privilege on a specific object at a time.
  ALL [ PRIVILEGES ]	Grants all privileges, except OWNERSHIP, on the compute pool.

Compute pool maintenance¶

As part of routine internal infrastructure maintenance, Snowflake regularly updates older compute pool nodes to ensure optimal performance and security. This includes operating system upgrades, driver enhancements, and security fixes. Maintenance involves replacing outdated nodes with updated ones every few weeks, with each node active for up to a month.

How services are scheduled on a compute pool¶

At the time of creating a service, you might choose to run multiple instances to manage incoming load. Snowflake uses the following general guidelines when scheduling your service instances on compute pool nodes:

• All containers in a service instance always run on a single compute pool node. That is, a service instance never spans across multiple nodes.

• When you run multiple service instances, Snowflake may run these service instances on the same node or different nodes within the compute pool. When making this decision, Snowflake considers any specified hard resource requirements (such as memory and GPU) as outlined in the service specification file (see containers.resources field).

  For example, suppose each node in your compute pool provides 8 GB of memory. If your service specification includes a 6-GB memory requirement, and you choose to run two instances when creating a service, Snowflake cannot run both instances on the same node. In this case, Snowflake schedules each instance on a separate node within the compute pool to fulfill the memory requirements.

Default compute pools for Notebooks¶

Starting release 8.46, Snowflake automatically provisions two compute pools in each Snowflake account (except the trial accounts) for running Notebook apps. These compute pools are exclusively for running Notebooks and cannot be used to create a Snowpark Container Services service.

• Compute pool name (as it appears in the Snowsight UI): SYSTEM_COMPUTE_POOL_GPU

  • Instance family: GPU_NV_S (see Instance family table)

  • Default configuration:

    • MIN_NODES=1

    • MAX_NODES=25

    • INITIALLY_SUSPENDED=true

    • AUTO_SUSPEND_SECS=600

• Compute pool name (as it appears in the Snowsight UI): SYSTEM_COMPUTE_POOL_CPU

  • Instance family: CPU_X64_S

  • Default configuration:

    • MIN_NODES=1

    • MAX_NODES=5

    • INITIALLY_SUSPENDED=true

    • AUTO_SUSPEND_SECS=600

The default configuration properties mean the following:

• Compute pools are initially in a suspended state and only begin incurring costs when Notebooks are started within them.

• If no Notebooks are running, these compute pools are automatically suspended after 10 minutes. Note the following:

  • By default, Notebooks are suspended automatically after 30 minutes of inactivity. After all Notebooks stop running on a default compute pool, Snowflake suspends the pool after 10 minutes.

  • To modify the auto-suspension policy for default compute pools, use the ALTER COMPUTE POOL SET AUTO_SUSPEND_SECS command. You can also adjust the Notebook auto-suspension policy. For more information, see Idle time and reconnection.

Default compute pools are provided for convenience. While any role in a Snowflake account can create a Notebook, only the ACCOUNTADMIN role is authorized to create compute pools. By using default compute pools, users can create Notebooks without needing an account administrator to configure a compute pool.

These compute pools are dedicated to Notebook workloads, and you can associate budgets with default compute pools to manage Notebook costs.

Note the following about the default compute pool permission:

• In a Snowflake account, the ACCOUNTADMIN role owns these compute pools. Administrators have full control over the compute pools, including modifying their properties, suspending operations, and monitoring consumption. If the default compute pools created by Snowflake are not needed, the ACCOUNTADMIN role can delete them. For example:

  USE ROLE ACCOUNTADMIN;
  ALTER COMPUTE POOL SYSTEM_COMPUTE_POOL_CPU STOP ALL;
  DROP COMPUTE POOL SYSTEM_COMPUTE_POOL_CPU;
  

  Copy

• By default, the USAGE permission on default compute pools is granted to the PUBLIC role, allowing all roles in the account to use them. However, the ACCOUNTADMIN can modify these privileges to restrict access if necessary.

  To restrict access to default compute pools to specific roles in your account, use the ACCOUNTADMIN role to revoke the USAGE permission from the PUBLIC role and grant it to the desired role(s). For example:

  USE ROLE ACCOUNTADMIN;
  REVOKE USAGE ON COMPUTE POOL SYSTEM_COMPUTE_POOL_CPU FROM ROLE PUBLIC;
  GRANT USAGE ON COMPUTE POOL SYSTEM_COMPUTE_POOL_CPU TO ROLE <role-name>;
  

  Copy

Guidelines and limitations¶

• CREATE COMPUTE POOL permission: If you cannot create a compute pool under the current role, consult your account administrator to grant permission. For example:

  GRANT CREATE COMPUTE POOL ON ACCOUNT TO ROLE <role_name> [WITH GRANT OPTION];
  

  Copy

  For more information, see GRANT <privileges>.

• Per account limit on compute pool nodes. The number of nodes you can create in your account (regardless of the number of compute pools) is limited to 120. In addition, there is a limit on the number of nodes allowed for each instance family (see the Node limit column in the instance family table). If you see an error message like Requested number of nodes <#> exceeds the node limit for the account, you have encountered these limits. For more information, contact your account representation.