Finding and fixing what went wrong when a data load fails

Troubleshooting Snowflake data ingestion means using the right metadata views and functions to see why a COPY INTO or Snowpipe load failed, was skipped, or loaded unexpected data. Snowflake exposes load status through COPY_HISTORY, LOAD_HISTORY, and pipe-specific functions rather than traditional logs, so knowing where to look and what each source retains is essential for diagnosis.

1 · Learn the must-know

  • Use the INFORMATION_SCHEMA.COPY_HISTORY table function (or ACCOUNT_USAGE.COPY_HISTORY for longer retention) to see file-level load status, row counts, and error messages for both COPY INTO and Snowpipe loads.
  • SYSTEM$PIPE_STATUS() reports a pipe's execution state, pending file count, and last error, which is the first place to check when Snowpipe seems stalled.
  • The VALIDATE function or VALIDATION_MODE copy option lets you preview which rows/files would fail before actually loading data, without writing to the table.
  • Snowflake tracks loaded file names/checksums for 64 days per table (load metadata), so a previously loaded file with the same name/path is skipped unless FORCE=TRUE is specified.
  • ON_ERROR (CONTINUE, SKIP_FILE, SKIP_FILE_<n>, ABORT_STATEMENT) determines whether a batch load stops or continues on row errors, and controls how much of the file is skipped versus loaded.
  • For Snowpipe specifically, cloud provider notification failures, missing SQS/Event Grid/Pub-Sub permissions, or stage/notification integration misconfiguration are common root causes and are visible via PIPE_USAGE_HISTORY and pipe error notifications (if configured).

3 · Keep going

Ready for more? Take a weighted mock or try free practice questions.

These are independent practice questions, written against this certification's published exam guide. They are not the certification vendor's own questions, and not the real exam.