采用高性能架构的 Snowpipe Streaming 的限制与注意事项¶

本文档概述了采用高性能架构的 Snowpipe Streaming 的已知限制及关键注意事项。

常规及服务级别限制¶

The service is available in all Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform (GCP) regions except for government-specific regions and regions in China.

每个管道的通道数：默认情况下，单个管道最多可以有 2,000 个活动通道。如果您的用例需要更多通道，请联系 Snowflake 支持部门。
Pipes for Snowpipe Streaming: The maximum number of PIPE objects configured for Snowpipe Streaming is limited to 1,000 per account and 10 per table. If you require more pipes, contact Snowflake Support.

每个通道具有以下软限制。如果您的应用程序需要更高的每通道吞吐量，请联系 Snowflake 支持部门以讨论提高这些限制。

在采用高性能架构的 Snowpipe Streaming 中，ON_ERROR 选项仅支持 CONTINUE。
数据吞吐量突然激增时，可能会出现短暂的端到端延迟增加，这是因为服务需弹性扩展以支持新的吞吐量级别。
不支持分区的 Iceberg 表。
MATCH_BY_COLUMN_NAME is not supported with default, auto-increment, or identity columns:

The MATCH_BY_COLUMN_NAME option isn't supported when you load data into tables that contain columns that are defined with the DEFAULT, AUTOINCREMENT, or IDENTITY properties. When you use this option, the streaming ingestion process explicitly inserts NULL values for these columns, overriding the intended default value or the auto-generation mechanism.

Workaround: To use these column properties, you must omit MATCH_BY_COLUMN_NAME. Instead, you define the pipe by using a COPY INTO statement that explicitly lists only the columns for which the source data provides values. The columns with the auto-generation properties must be omitted from the target column list to ensure that the table engine applies the defined value generation logic.