Dan Burkert commented on KUDU-2133:

Probably related to KUDU-1179.  Would be really nice to fuzz test the decoding 
of range partition rows, since it should be considered un-trusted input.

> Master Server Crash from mismatched schemas using Spark Context
> ---------------------------------------------------------------
>                 Key: KUDU-2133
>                 URL: https://issues.apache.org/jira/browse/KUDU-2133
>             Project: Kudu
>          Issue Type: Bug
>          Components: java, rpc
>    Affects Versions: 1.4.0
>            Reporter: Benjamin Guidarelli
>            Priority: Major
>         Attachments: 3b055209-58c0-24a8-3cd396a4-4fa7d010.dmp, 
> 6b08ef81-ddb8-8caa-241e8ae7-7800226c.dmp
> While creating a table in Kudu using kud-spark KuduContext I omitted the 
> `.nullable(true)` option from several fields when I generated a schema so I 
> could add partitions using schema.newPartialRow.  KuduContext.createTable 
> takes a spark StructType as an argument and adds the nullable(true).  When 
> calling createTable the MasterServer reliably crashes, I think, because of 
> this conflict in nullable fields. 

This message was sent by Atlassian JIRA

Reply via email to