[Bug] IcebergScanNode throws exception "Split size must be > 0: 0"

[wenwj0]

Search before asking

• I had searched in the issues and found no similar issues.

Version

Doris 2.1.6

What's Wrong?

After I upgrade doris from 2.1.5 to 2.1.6, I occasionally encountered an exception when querying iceberg table.

The exception may be thrown in org.apache.iceberg.util.TableScanUtil

public static CloseableIterable<FileScanTask> splitFiles(CloseableIterable<FileScanTask> tasks, long splitSize) {
        Preconditions.checkArgument(splitSize > 0L, "Split size must be > 0: %s", splitSize);
        Iterable<FileScanTask> splitTasks = FluentIterable.from(tasks).transformAndConcat((input) -> {
            return input.split(splitSize);
        });
        return CloseableIterable.combine(splitTasks, tasks);
    }

What You Expected?

Expect no exeception

How to Reproduce?

QuerySQL is

SELECT * FROM 
doris_hive_xxx.xxx.ods_xxx_event_iceberg_day 
LIMIT 10;

Anything Else?

No response

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct

[wenwj0]

@morningman PTAL, I guess it maybe related to #38311

[sacredcodex]

Did you solve this problem? I faced it when selecting the logical view of Iceberg.

[wenwj0]

Did you solve this problem? I faced it when selecting the logical view of Iceberg.

@sacredcodex No, I change to use doris-2.1.5.
By the way, do you config hadoop kerberos? I saw some error logs related to that.

[sacredcodex]

Thank you for your reply. We used Doris 2.1.2 before and config Kerberos then. We recently upgraded to 2.1.5, and we met with this problem. It might not be caused by Kerberos. Thank you anyway.

谢谢您的回复，我们之前使用的doris2.1.2，并且配置过Kerberos了。我们最近升级到了2.1.5版本并遇到了这个问题。应该不是Kerberos导致的，再次感谢您的帮助。

[morningman]

You can try

1. set enable_fallback_to_original_planner=false;
   run sql again to see if there are errors.