You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Martijn Visser (Jira)" <ji...@apache.org> on 2022/01/04 19:10:00 UTC

[jira] [Updated] (FLINK-25106) Allow specifying specific offsets for each partition for Upsert-Kafka connector

     [ https://issues.apache.org/jira/browse/FLINK-25106?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Martijn Visser updated FLINK-25106:
-----------------------------------
    Description: 
Currently, FLINK's "kafka" connector ignores all the tombstone messages, whereas the "upsert-kafka" connector supports tombstone messages and sets the type of the row to RowKind.DELETE ([Code Link|https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-kafka/src/main/java/org/apache/flink/streaming/connectors/kafka/table/DynamicKafkaDeserializationSchema.java#L126]).


I think that the feature need is actually an extension on the upsert-kafka connector (to support specific offsets), because the key difference there is between the kafka connector and the upsert-kafka connector is how tombstones are dealt with currently.

  was:
Currently, FLINK's "kafka" connector ignores all the tombstone messages, whereas the "upsert-kafka" connector supports tombstone messages and sets the type of the row to RowKind.DELETE ([Code Link|https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-kafka/src/main/java/org/apache/flink/streaming/connectors/kafka/table/DynamicKafkaDeserializationSchema.java#L126]).

I wanted to know if it is feasible to support tombstone messages in "kafka" connector by setting all the value fields to NULL and the RowKind to DELETE. I could also raise a PR with the respective changes if required.


> Allow specifying specific offsets for each partition for Upsert-Kafka connector
> -------------------------------------------------------------------------------
>
>                 Key: FLINK-25106
>                 URL: https://issues.apache.org/jira/browse/FLINK-25106
>             Project: Flink
>          Issue Type: Improvement
>          Components: API / Core
>            Reporter: Varun Yeligar
>            Priority: Minor
>
> Currently, FLINK's "kafka" connector ignores all the tombstone messages, whereas the "upsert-kafka" connector supports tombstone messages and sets the type of the row to RowKind.DELETE ([Code Link|https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-kafka/src/main/java/org/apache/flink/streaming/connectors/kafka/table/DynamicKafkaDeserializationSchema.java#L126]).
> I think that the feature need is actually an extension on the upsert-kafka connector (to support specific offsets), because the key difference there is between the kafka connector and the upsert-kafka connector is how tombstones are dealt with currently.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)