[ORACLE] XMLTYPE + LOBs + Generated Columns

[Vitor Pereira]

Hi,

I hope this message finds you well.

We have another problem related to the PostSelector. As of now, we're using version 3.0.8, although there is one problem:

If the table has an XMLTYPE column and a generated column too, and the sink table is created via python, the sink connector displays an error regarding schema evolution (we're using ‘schema.evolution’: ‘none’, since we want to do all the DDL externally).

We ran some tests:

• Test1: Table with XMLTYPE + generated column - Failed
• Test2: Table with CLOB + XMLTYPE - Worked (we selected the special types and created the table with python. Everything as it was supposed to)
• Test3: Table with CLOB + Generated column - Worked

We also ran some queries in the ALL_TAB_COLS view and there are inconsistencies in COLUMN_ID and SEGMENT_COLUMN_ID.

Could this be related to [1], [2], [3] and [4]?

Is there any way to solve this problem? If you need more information, let me know.

Best regards

[1] https://issues.redhat.com/browse/DBZ-8579

[2] https://issues.redhat.com/browse/DBZ-8597

[3] https://issues.redhat.com/browse/DBZ-8673

[4] https://issues.redhat.com/browse/DBZ-7835

[Chris Cranford]

Hi Vitor -

That sounds like a potential bug. For Test1 that should ultimately get mapped to a CLOB + generated column just like Test3, so why it fails seems strange.

Could you raise a Jira issue including the following:

    * DDL of the source table
    * Screenshot/data dump of the inconsistencies in ALL_TAB_COLS for the table
    * TRACE logging for both the source & JDBC sink connector for this use case
    * Connector configurations for both source and sink

TIA,
-cc

--
You received this message because you are subscribed to the Google Groups "debezium" group.
To unsubscribe from this group and stop receiving emails from it, send an email to debezium+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/debezium/eb076f4c-a7a8-4d24-82ca-e40a4ee45de3n%40googlegroups.com.

[Chris Cranford]

I failed to mention, could you also provide the DDL for the sink table too that you created using python on the issue as well.  Thanks!

[Vitor Pereira]

Hi, Chris.

Fortunately there is no bug. If the type is propagated through "datatype.propagate.source.type" everything works as supposed to.

Thanks,

Vitor Pereira