Using Snowflake stage volumes with services¶

Note

Using Snowflake stage volumes with services is currently not available to accounts in Google Cloud.

Snowflake supports various storage volume types for your application containers, including internal stage, local storage, memory storage, and block storage volumes. This section explains how to configure volumes and volume mounts for internal stages.

Stage volumes provide your service with access to internal stages by using FileSystem APIs, simplifying your code compared to using a Snowflake driver and GET and PUT SQL commands.

Stage volumes can also be more performant for scenarios with streaming reads or writes of large data files. Stage volumes work best when the application uses the stage volume as a convenient alternative API for stages, rather than expecting a fully capable file system.

There are currently two implementations of stage volumes: the generally available version and a newer version that is currently in preview.

About the new stage volume implementation¶

The generally available stage volume implementation uses a shared cache for reads and writes. Although this works well for some scenarios, you can’t control whether data is read from the cache or directly from the stage, which might not be suitable for all applications. Additionally, when you use the cache for reads and writes, this can introduce performance overhead.

The new stage volume implementation, currently in preview, uses only limited in-memory caching. This provides more predictable behavior and significantly faster throughput. This version will eventually replace the current generally available implementation. Snowflake recommends evaluating this preview version unless your workload requires random writes or file appends, which aren’t currently supported.

Keep in mind the following additional considerations when you use the new stage volume implementation:

It is optimized for large, sequential reads and writes, providing strong performance for these access patterns. For best results, read and write data in large, contiguous chunks.
Random read performance might be lower with the new implementation because of the removal of caching. However, without a local disk cache, consistency across volumes is improved. File changes are written directly to the backing stage when the file is closed, with no local disk buffering.
Reads always return the latest data, making this configuration better for sharing data between services.
Random writes or file appends isn’t supported. Attempting these operations results in an error.

Specify a Snowflake stage volume in a service specification¶

To create a service where service containers use a Snowflake stage volume, specify the required settings in the service specification as shown in the following steps:

To specify the stage volume, use the spec.volumes field.
- Use the generally available version of the stage volume implementation, as shown in the following example:
  volumes: - name: <name> source: <stage_name>
  Copy
  The following fields are required:
  - name: Name of the volume.
  - source: The Snowflake internal stage or folder on the stage to mount, for example @my_stage, @my_stage/folder. Quotes must surround this value.
- Use the new preview version of the stage volume implementation, as shown in the following example:
  volumes: - name: <name> source: stage stageConfig: name: <stage_name>
  Copy
  The following fields are required:
  - name: Name of the volume.
  - source: Identifies the type of the volume (stage).
  - stageConfig: Identifies the Snowflake internal stage or folder on a stage to mount, for example @my_stage, @my_stage/folder, or @my_db.my_schema.my_stage/folder/nestedfolder. Double quotes must surround this value.
Use the spec.containers.volumeMounts field to describe where to mount the stage volume in your application containers, as show in the following example:
```
volumeMounts:
- name: <name>
  mountPath: <absolute_directory_path>
```
Copy
The information you provide in this field is consistent across all supported storage volume types and applies to both implementations of stage volumes.

Example¶

This examples illustrates the difference between mounting a Snowflake stage by using the existing and the new stage volumes. In the example service specification, the app container mounts an internal stage @model_stage by using both the existing and new stage volume implementations.

@model-legacy stage volume configuration directs Snowflake to use the generally available implementation of the stage volume.
@model-new stage volume configuration specifies the stageConfig field that directs Snowflake to use the preview implementation of the stage volume.

spec:
  containers:
  - name: app
    image: <image1-name>
    volumeMounts:
    - name: models-legacy
      mountPath: /opt/model-legacy
    - name: models-new
      mountPath: /opt/model-new
  volumes:
  - name: models-legacy
    source: "@model_stage"
  - name: models-new
    source: stage
    stageConfig:
      name: "@model_stage"

Copy

The volumeMounts field specifies where inside the container to mount the stage volume. This specification remains same for both the stage volume implementations.

Access control requirements¶

The service’s owner role is the role that is used to create the service. It is also the role the services use when interacting with Snowflake. This owner role determines the permissions granted to application containers for accessing a mounted stage. The owner role must have the READ privilege on the stage.

If the owner role doesn’t have the WRITE privilege on a stage, the mount for that stage is read-only. That is, the containers can only read the files from the stage. The owner role needs the WRITE privilege on a stage for the stage mount to support both read and write.

Guidelines when using stage volumes¶

This section provides you with guidelines to follow when you implement application code in which containers use stage volumes.

Common guidelines for both implementations of stage volumes¶

Stage mount is optimized for sequential reads and writes.
Stage mount I/O operations might have higher latencies than I/O operations on the container’s file system and block storage volumes. You should always check the status code of I/O operations to ensure they succeeded.
Stage mounts upload file updates asynchronously. Changes to files on a stage mount are only guaranteed to be persisted to the stage after the file descriptor is successfully closed or flushed. There might be a delay before the changes to files on a stage mount become visible to other containers and Snowflake.
Each directory in a mounted stage should contain fewer than 100,000 files. Expect readdir latency to increase with the number of files in the directory.

Guidelines when using the generally available version of the stage volume implementation¶

Avoid concurrently writing to multiple files within a stage mount.
Stage mount isn’t a network file system. Don’t use stage mounts for multi-client coordination.
Don’t open multiple handles to the same file concurrently. Use opened file handles for either read or write operations, but not a mixture of both. To read from a file after writing to it, close the file and then re-open the file before reading.

Guidelines when using the new preview version of the stage volume implementation¶

Concurrent writes to the same file from multiple stage mounts — same stage volume mounted on different containers — aren’t recommended.
The absence of a local disk cache improves consistency across mounts. File changes are flushed directly to the backing stage upon closing the file, with no local disk buffering. Reads always return the latest data, making the new stage mount better for sharing data between services.
Read and write data in large, contiguous chunks for optimal performance. The performance penalty for small reads and writes when compared to the generally available stage volume implementation, can mitigate the performance gains from the new implementation.

Limitations when using stage volumes¶

This section describes limitations you should be aware of when you implement application code in which containers use stage volumes. If you encounter any issues with these limits, contact your account representative.

Common limitations for both implementations of stage volumes¶

You can only mount a stage or a subdirectory in a stage; for example, @my_stage, @my_stage/folder. You can’t mount a single file in a stage; for example, @my_stage/folder/file.
External stages aren’t supported. Only Snowflake internal stages are supported.
A maximum of 5 stage volumes is allowed per service. For more information, see spec.volumes.
A maximum of 8 stage volumes per compute pool node are supported. Snowflake manages the stage mount per node limit similar to how it manages memory, CPU, and GPU. Launching a new service instance can cause Snowflake to launch new nodes when no existing nodes have the capacity to support the requested stage mounts.
Stage mounts are not fully POSIX compatible file systems. For example:
- File and directory renames are not atomic.
- Hard links are not supported.
The Linux kernel subsystem inode notify (inotify) that monitors changes to file systems doesn’t work on stage mounts.

Limitations when using the generally available version of the stage volume implementation¶

The stage volume capabilities vary depending on the cloud platform for your Snowflake account:
- Accounts on AWS support internal stages with both SNOWFLAKE_FULL and SNOWFLAKE_SSE stage encryption. For more information, see Internal stage parameters.
- Accounts on Azure currently support internal stages with SNOWFLAKE_SSE encryption. When you run CREATE STAGE, use the ENCRYPTION parameter to specify the encryption type: CREATE STAGE my_stage ENCRYPTION = (TYPE = 'SNOWFLAKE_SSE');
- Accounts on Google Cloud aren’t supported.
Concurrent writes to the same file from multiple stage mounts — that is, the same stage volume mounted on different containers — aren’t supported.

Limitations when using the new preview version of the stage volume implementation¶

The stage volume capabilities vary depending on the cloud platform for your Snowflake account:
- Accounts on AWS support internal stages with both SNOWFLAKE_FULL and SNOWFLAKE_SSE stage encryption (see Internal stage parameters).
- Accounts on Azure and Google Cloud currently support internal stages with SNOWFLAKE_SSE encryption. When you run CREATE STAGE, use the ENCRYPTION parameter to specify the encryption type: CREATE STAGE my_stage ENCRYPTION = (TYPE = 'SNOWFLAKE_SSE');
Random writes, and file appends aren’t supported.
Each stage that is mounted requires 512 MB memory per stage. This means that there is a limitation on the number of stage volumes that can be used based on instance size. Mounting the volume on multiple containers doesn’t increase memory consumption.