A Snowflake Notebook is being used to train a machine learning model. The notebook contains multiple cells: one to load data, one to perform feature engineering, and one to train the model using Snowp

Question

Options

• AThe notebook session timed out due to inactivity. Increase the SESSION_IDLE_TIMEOUT setting
• BThe Snowflake warehouse being used for the notebook session is undersized, leading to memory
• CThe Snowpark ML library encountered an incompatibility issue with the current Snowflake version.
• DThe feature engineering process resulted in an excessively large intermediate dataset that
• EThe notebook is configured to automatically save checkpoints. This feature is causing overhead

• A
  4% (2)
• B
  80% (43)
• C
  11% (6)
• E
  6% (3)

Explanation

With ~50 GB of data and an unexplained kernel restart (no error message), the two most likely causes are memory-related: (B) An undersized virtual warehouse lacks sufficient memory to handle the volume of data, causing the Snowpark session worker to be killed by the OS OOM (out-of-memory) killer - the kernel restarts automatically with no explicit error displayed; and (D) The feature engineering step produced a large intermediate dataset (e.g., due to a join explosion, unpivoting, or collecting data to the driver) that exceeded available memory, triggering the same OOM crash. Option A (idle timeout) is unlikely since the notebook was actively processing. Option C (library incompatibility) would typically surface a version mismatch error rather than a silent restart. Option E (checkpoint overhead) is not a standard Snowflake Notebook behavior and would not cause a kernel restart. The remedy for B is to use a larger warehouse; for D, optimize transformations to avoid materializing large intermediate results (e.g., push computation to Snowflake via Snowpark lazy evaluation).

No community discussion yet for this question.