[
https://issues.apache.org/jira/browse/FLINK-20060?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ASF GitHub Bot updated FLINK-20060:
-----------------------------------
Labels: auto-deprioritized-major auto-deprioritized-minor
pull-request-available (was: auto-deprioritized-major auto-deprioritized-minor)
> Add a Collector to KinsesisDeserializationSchema
> ------------------------------------------------
>
> Key: FLINK-20060
> URL: https://issues.apache.org/jira/browse/FLINK-20060
> Project: Flink
> Issue Type: New Feature
> Components: Connectors / Kinesis
> Reporter: Timo Walther
> Priority: Not a Priority
> Labels: auto-deprioritized-major, auto-deprioritized-minor,
> pull-request-available
>
> We did not add support for a collector in the KinesisDeserializationSchema.
> The problem with that connector lays in the threading model, where there is a
> pool of threads that read and deserialize records and then they handover the
> deserialized messages through a queue to the main thread. The problem is that
> we would need to create many temporary collections to put the deserialized
> records into the handover queue, which potentially would significantly affect
> performance, especially in the usual case of deserializing a single record
> from a single message.
> This means that we can currently not support the Debezium format for in the
> SQL connector if the Debezium record needs to emit 2 rows (UPDATE_BEFORE and
> UPDATE_AFTER).
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
