Hi everyone, I'm running into a "NULL value in non-nullable column" error while loading data into StarRocks from a Parquet file. I verified the Parquet file and confirmed that the column does not contain any NULL values. However, during the load process, StarRocks still reports that a NULL value is being inserted into a non-nullable column. Has anyone encountered this issue before? Are there any known reasons why StarRocks might interpret a non-NULL value as NULL during a Parquet load? Could this be related to schema mismatches, data type conversions, column mapping, or another loading behavior? Any suggestions on how to debug this would be greatly appreciated. Thanks!

Hi @Mahesh Thanks for the report. Could you share a few details so we can narrow this down and try to reproduce it on our side? 1. Full load statement and StarRocks version — Broker Load, INSERT FROM FILES, or Stream Load, including any

columns

mapping or

SET

clauses. 2. Exact error message with the

Row: [...]

part — the position of the NULL in that row shows which column and stage produced it. 3. CREATE TABLE statement and the Parquet file itself — with these two we can reproduce the load directly. Please share the original file that triggered the error, not a sampled or trimmed copy. This error is often caused by specific values failing type conversion, so a partial file may not contain the offending rows and the issue may not reproduce. 4. Quick isolation test — run

SELECT * FROM FILES(...)

on the same file and check whether StarRocks itself reads NULLs for that column. If NULLs appear, it's a schema or type mapping issue. If not, the NULL is generated during conversion in the load pipeline. - FILES | StarRocks If the file or DDL is difficult to share here on Slack, feel free to send them to kjjs1996@gmail.com instead.